The Inflation Number the New Fed Chair Stopped Looking At

On Thursday the Federal Reserve's favorite inflation gauge printed a number that looks like an emergency. The Personal Consumption Expenditures price index, the Fed's preferred inflation yardstick, rose 4.1% over the year through May, a three-year high. Within hours the hawks had their headline. Bank of America, which a week earlier expected no rate moves at all in 2026, reversed itself, called for three rate hikes by December, and declared the Fed's inflation problem 'unambiguously worse.'

But sitting in the very same data release is a second number that says something close to the opposite. And it is, by his own repeated testimony, the number the Fed's new chairman trusts most.

Forget, for a moment, whether 4.1% is high. The only question that decides whether the Fed should raise rates is this: is May's jump a broad inflation that has spread across the economy, or a narrow spike concentrated in a few shock-hit categories, gasoline from the 2026 Iran war, and the last of the tariff pass-throughtariff pass-throughThe share of an import tax that gets passed on to consumers as higher prices rather than absorbed by sellers.?

This is not a nuance. It is the whole case. You raise interest rates to cool an economy generating inflation everywhere, from haircuts to rent to restaurant meals. You do not raise rates to fight a one-off jump in oil and imported goods, because tightening into a supply shocksupply shockA sudden disruption to the supply of goods, like a war spiking oil prices, that pushes prices up for reasons unrelated to how hot demand is. adds a second injury: you slow hiring to punish the economy for a price increase it did not cause and cannot undo.

'How broad is it?' is the difference between sound policy and a costly mistake.

Three events in nine days set the stage. On 17 June, at his first meeting as chair, Kevin Warsh held the policy rate steady at 3.50%-3.75% on a unanimous vote, but the Fed's 'dot plotdot plotA grid the Fed publishes showing where each official privately thinks interest rates should be in coming years, a peek at the committee's leanings.', the grid of where officials think rates should go, flipped hawkish, with the median end-2026 dot jumping to 3.8% from 3.4% and nine of eighteen members now penciling in a hike before year-end. Warsh, tellingly, declined to submit a dot of his own.

On 22 June, Bank of America abandoned its no-hikes call and projected three quarter-point increases, in September, October and December, to a 4.25%-4.5% range, citing inflation that is 'unambiguously worse.' Then on 25 June the Bureau of Economic Analysis published the May PCE report: headline 4.1%, a three-year high. The hawkish case had its number.

Warsh himself is new to the chair, sworn in on 22 May as the Fed's 17th chairman, after the narrowest Senate confirmation vote in the institution's history, 54-45. The pivot toward hikes is the first major signal of his tenure.

There is a clean tool for the breadth question, and it is not a matter of opinion. It is the Dallas Fed Trimmed Mean PCETrimmed Mean PCEA Dallas Fed inflation gauge that throws out the categories with the most extreme price changes each month, both the fastest risers and the fastest fallers, and measures only the calm middle, to reveal broad underlying inflation.. Picture all the hundreds of things people buy, sorted each month from the fastest price gainers to the fastest fallers. The trimmed mean throws out the extremes at both ends and measures only what the calm middle is doing. It is the inflation rate after you ignore whatever is screaming loudest in either direction.

If inflation is genuinely broad, the middle is hot and the trimmed mean climbs. If a spike is concentrated, the trim removes the offending categories and the trimmed mean stays put. So the gap between the headline and the trim is, quite literally, a breadth meter. In May it read 1.7 percentage points, headline 4.1%, trimmed mean 2.4%, and the trim was easing, not rising. That is the fingerprint of a few categories doing the damage, not the whole basket.

Decompose the headline and the same story falls out of simple arithmetic. Start from the broad middle at 2.4%. Add back the food and energy the core excludes, headline minus core is 0.7 of a point. Add back the extreme core movers the trim discards, core minus the trimmed mean is another point. Stack those three and you are at 4.1%. Roughly two-thirds of the distance from target to the scary headline is built from the pieces both the core and the trim are designed to look past.

The 2022 comparison sharpens the point. Back then the inflation was broad and demand-driven, and the trim showed it: in May 2022 the headline ran at 6.3% and the trimmed mean itself sat at 4.0% and rising. The middle was on fire. Today the headline is lower, and the middle is barely above target and cooling. Same alarm bell on the headline; opposite reading underneath.

What is screaming loudest is not a mystery. Since late February, Iran's near-total blockade of the Strait of HormuzStrait of HormuzA narrow sea passage at the mouth of the Persian Gulf through which roughly a fifth of the world's traded oil moves; blocking it chokes global oil supply and pushes prices up., the narrow sea lane that carries about a fifth of the world's traded oil, has throttled global crude supply, and US pump prices climbed by more than a dollar a gallon at the worst of it. That is a textbook supply shock: prices up for reasons that have nothing to do with how hot American demand is. By late June, with Iran moving to reopen the strait, Brent had slid back toward $76 a barrel and the pressure was fading.

The second pusher is tariffs. As import taxes worked their way onto shelf prices, they lifted the cost of affected goods, but as a one-time step up in the level of prices, not a permanent change in their growth rate. Dallas Fed economists Ron Mau and Tucker Smith found that the effect of realized tariff changes on PCE prices peaked in the first quarter of 2026. A level shift that has already crested is exactly the kind of move the trimmed mean is engineered to set aside: it shows up as a cluster of extreme gainers, gets trimmed off, and leaves the broad middle untouched.

So the two forces lifting the headline, war-era oil and crested tariff pass-through, are precisely the categories the breadth gauge discards by design. Remove them and you are left with an underlying rate near target. That is the disinflation-camp case in one sentence.

Here is the awkward part for the new chairman. Kevin Warsh did not adopt the trimmed mean for this article's convenience; he has championed it for months. At his confirmation hearing he dismissed the standard inflation gauge as only a 'rough swag' of real price pressure and named the Dallas Fed trimmed mean as his preferred measure. 'What I'm most interested in is what's the underlying inflation rate,' he said, 'not what's the one-time change in prices because of a change in geopolitics or a change in beef.' War-driven oil and a one-off tariff step are about as close to 'a change in geopolitics' as the data gets.

By his own test, May is the one-time change in prices he said he would look through, not the underlying rate he said he would target.

Yet the committee he now leads tilted toward hikes, and Warsh let it, while pointedly withholding his own dot. The pivot sits in visible tension with the metric he spent his confirmation defending. That does not make the hawks right or wrong; it means the chairman's favorite instrument and his committee's body language are pointing in opposite directions, and he has not yet said which one wins.

Intellectual honesty requires the strongest version of the other side, and it is not nothing. Strip out housing and the volatile stuff and look at core services, haircuts, restaurant meals, medical care, the so-called 'supercore.' It is the part of inflation that tracks wages rather than supply chains, and it is the truest tell of home-grown price pressure. By the May reading it was running near 3.9% on a yearly basis, well above the roughly 2.6% pace that would square with 2% inflation, and analysts at Employ America flagged it as a genuine problem for the Fed. If supercore is heating, the trim could be lulling everyone to sleep.

Two further cautions belong on the record. Some Fed economists, including Dallas Fed President Lorie Logan, have warned that an unusual lean in the spread of price changes is currently pulling the trimmed mean lower than the underlying trend would justify, the gauge may be flattered right now. And Mau and Smith's own work cautions that a trimmed mean sitting below core is not, by itself, strong evidence that disinflation is coming if tariff-driven price pressure keeps spreading. The trim is a breadth meter, not a crystal ball.

But the engine that turns a services blip into a wage-price spiral is labor costs, and that engine is idling. The Employment Cost Index, the cleanest read on what employers pay in wages and benefits, rose about 3.4% over the year through the first quarter, holding roughly steady rather than accelerating. Hot services prices without accelerating wages behind them look like a shock still passing through, not a spiral taking hold. The trim, on this read, looks early rather than plainly wrong.

A reaction is not an effect. Bank of America's flip and the dot plot's hawkish tilt are market and committee reactions to a frightening headline; whether the economy actually needs tighter policy depends on the breadth question, and the breadth gauge says the pressure is narrow and easing. The two should not be confused.

From the current 3.50%-3.75% range, three paths are in play by year-end. The base case tracks the Fed's own median dot toward roughly one hike. The hawkish path is BofA's three increases, to 4.25%-4.5%, which makes sense only if breadth broadens and wages catch up. The dovish path, holding, is what Warsh's favorite metric, taken at face value, would counsel: let a fading supply shock fade. The coming months of supercore and wage data will decide which of his two instruments the chairman ends up believing.