The Bureau of Economic Analysis released May personal consumption expenditures data on Thursday. Three numbers, three different stories. Headline PCE inflation rose to 4.1 percent year-over-year, the highest since April 2023. Core PCE, which excludes food and energy, came in at 3.4 percent, the highest since October 2023. The Dallas Fed's trimmed mean PCE, which strips the most extreme price changes from both tails, read 2.4 percent.
The gap between core and trimmed mean is a full percentage point. In normal times, the two measures track closely. When they diverge, the distance between them is almost definitionally the contribution of extreme price moves. The outliers that one measure keeps, the other discards.
Kevin Warsh knows this. During his Senate confirmation hearing on April 21, the incoming Fed Chair told Congress which thermometer he trusts. "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." He named the Dallas Fed trimmed mean and the Cleveland Fed median PCE as his preferred instruments. Both strip outliers. Both were designed for exactly this kind of moment: supply shocks muddying the signal.
The Reading
The trimmed mean works by sorting every component of personal consumption expenditures by its monthly price change, then cutting 24 percent from the bottom and 31 percent from the top. What remains is the middle of the distribution, the prices moving at ordinary speeds. At 2.4 percent over twelve months, the trimmed mean is saying that if you remove the extreme movers, underlying inflation is within striking distance of the Fed's 2 percent target.
The components being trimmed are not a mystery. Energy prices surged after U.S. military operations against Iran drove oil above sustained highs. Import prices rose as tariffs added cost layers across consumer goods. These are the tail readings, the extremes that the trimmed mean was built to exclude. They are also, dollar for dollar, the gap between 2.4 and 3.4 percent.
Warsh's preferred thermometer is giving him the answer he described wanting: the underlying rate, stripped of geopolitics. It reads near target.
The Bind
He cannot use it. At his first FOMC meeting on June 17, Warsh presided over a hawkish shift. Nine of eighteen officials who submitted projections see at least one rate hike before year-end. Six see at least two. The median federal funds rate projection rose to 3.8 percent from 3.4 percent. Warsh's post-meeting statement was ten words on inflation: "Persistently high prices are a burden for the American people." The committee, he added, "will deliver price stability."
The policy response tracks core PCE at 3.4 percent, not trimmed mean at 2.4 percent. The thermometer driving decisions is not the one Warsh told Congress he preferred.
The reasons he cannot switch run deeper than preference. Brookings researchers warned in April that "switching to a new metric when inflation has been above the Federal Reserve's 2 percent PCE target for five years could be seen as moving the goal posts, potentially threatening the central bank's credibility." The Dallas Fed's own economists published a paper in April cautioning that the trimmed mean carries a downward bias when price changes become positively skewed, precisely the pattern produced by supply shocks like a war and a tariff regime. The Richmond Fed, in a May analysis titled "Core Confusion," noted that trimmed measures lagged headline and core during the 2021 inflation rise, missing early signals of broadening pressure.
Every argument against the trimmed mean is an argument that Warsh's stated analytical framework does not apply to the current environment. The instrument he chose was calibrated for a world where supply shocks are temporary and narrow. If the war persists, if tariffs entrench, the outliers stop being outliers. They become the new distribution.
The Error
Rate hikes treat demand. They make borrowing more expensive, slow spending, cool the economy. They do not make oil cheaper. They do not remove tariffs. They do not end wars. If core PCE at 3.4 percent is elevated primarily because of energy and import prices (and the one-percentage-point gap to the trimmed mean says it largely is), then rate hikes are a demand-side tool applied to a supply-side problem.
This is not a theoretical distinction. The Fed's own history contains the precedent. In 2022, the FOMC raised rates from near zero to 5.25 percent over sixteen months. Inflation fell, but whether it fell because of the rate hikes or because supply chains healed remains one of the most contested questions in macroeconomics. The trimmed mean's defenders argue that underlying demand-driven inflation never rose as high as core PCE suggested, and that the rate hikes were partially overtreating a supply shock.
Chicago Fed President Austan Goolsbee, historically among the most dovish FOMC members, said Thursday that core inflation was "still too high" and "trending the wrong way." When the dove agrees with the hawk, the policy direction is set. The question is not whether rate hikes come. The market has priced them. The question is what they fix.
The Gap
The one percentage point between 3.4 and 2.4 is the cost of the war and the tariffs, measured in inflation. It is also the distance between the thermometer Warsh told Congress he trusts and the thermometer he is acting on. If the trimmed mean is right, the Fed is about to tighten into a supply shock, raising rates against inflation it cannot reach. If the trimmed mean is wrong, if its downward bias during skewed distributions means it is understating true pressure, then core PCE is the honest reading and Warsh's confirmation testimony was an analytical framework that does not survive contact with his own economy.
Either way, the gap closes. If energy and tariff pressures ease, core PCE falls toward the trimmed mean and rate hikes become unnecessary. If they persist and broaden, the trimmed mean rises toward core PCE and Warsh's preferred instrument confirms the diagnosis he is already treating.
The thermometer he chose reads 2.4 percent. The one driving policy reads 3.4 percent. Somewhere in the space between them is the question of whether the next rate hike treats the disease or the symptom.