Core PCE Holds at 3.3% (July 2026): One Line Worth 2% of the Basket Delivered Half the Month — and Real Spending Went to Zero

July core PCE rose 0.246% on the month and 3.344% on the year — an annual rate that did not fall, against a consensus that said it would. The Cleveland Fed's nowcast had 0.247%. The number consensus missed by seven basis points was, almost exactly, one line item: portfolio management and investment advice prices jumped 5.624% in a single month and contributed roughly 0.12 percentage points, close to half the increase. And in the same report, real consumer spending stopped: up $1.3 billion, the first zero month since May 2023, with real goods spending down 0.61%.

This piece was published as a preview on the morning of 26 August, arguing that the seven-basis-point gap between consensus and the Cleveland Fed's model was not noise about a hundred categories but a disagreement about one, and that the resolution would be a measurement of asset values rather than of what households pay. The 8:30 a.m. print settled it. This update carries the verified outcome and what it changes.

What actually happened

Every figure below is from the Bureau of Economic Analysis's Personal Income and Outlays release for July 2026 and the NIPA price tables published with it. Unrounded figures are computed from BEA's own published index levels, which is the only way to see a seven-basis-point argument settled.

Two things deserve to be read slowly. The first is the accuracy of the nowcast: 0.247% projected against 0.246% delivered on core, and 0.155% against 0.156% on headline. That is a tenth of a basis point on both, in a month when the consensus of forty-odd forecasters was seven basis points away. The second is that the annual core rate did not move at all. BEA updated its estimates for April through June in this release; on the vintage published today, June reads 3.344% and July reads 3.344%. The word for that is not "eased".

The line that did it

The preview argued that portfolio management and investment advice was the disputed category. BEA's underlying detail table for PCE prices shows exactly how much of the month it was.

Portfolio management is $418.6bn at a seasonally adjusted annual rate — 1.9% of total PCE, and roughly 2.1% of the core basket once food and energy are removed. A 5.624% monthly price increase on a 2.1% weight contributes approximately 0.12 percentage points. The core index rose 0.246 points. One category worth two cents in the dollar delivered close to half the month.

Take it out and the arithmetic is unambiguous: the remaining 98% of the core basket rose about 0.13% — below the 0.18% consensus for the index as a whole. The month that beat expectations was, outside this line, a month that undershot them.

There is a second, independent way to see the same thing. BEA publishes a supplemental market-based measure that excludes prices which are imputed rather than observed in a transaction. Market-based core PCE rose 0.151% in July against the full measure's 0.246%, and runs at 3.025% year over year against 3.344%. The memo row in the table above shows why: on the market-based basis, financial services and insurance prices fell 1.596% in the same month the published index has them rising 1.170%. Two different methods, one difference of about a tenth of a point, both pointing at the same category.

The July paradox, confirmed

The strangest part of the preview was a prediction about direction that had nothing to do with forecasting inflation. July 2026 was a down month for the assets people associate with the fee base: the Philadelphia Semiconductor Index fell 20.6% over the calendar month and the Nasdaq 100 fell 6.6%. Yet the fee price index rose 5.6%.

The mechanism is the lag in the base, not a contradiction. Managed money is billed ad valorem — a percentage of assets under management — and the balances those July invoices were struck against were accumulated through a second quarter in which the SOX posted its best quarter since the index began in 1994 and the Nasdaq 100 gained 28%. The producer price index treats the resulting revenue-per-unit increase as a price increase. So the statistic recorded the melt-up while the drawdown was happening.

The forward implication is a timing statement rather than a forecast, and it runs the same way in reverse. Whatever July's equity drawdown does to this line, it does with a lag — and it will do it into a rewritten deflator.

The other half of the report: income up, spending flat

The inflation number will get the headlines. The spending number is the one that changes what the report says about the economy.

Personal income rose $115.1bn, or 0.4% — double June's rate and above the 0.2% consensus. Disposable income rose 0.5%, and real disposable income rose 0.4%. Households had more purchasing power in July than in June, and it was not an inflation illusion.

They did not spend it. Real PCE rose $1.3bn, which BEA reports as less than 0.1% and which is the first month at zero since May 2023. The split underneath is sharper than the total:

In nominal terms, services spending rose $86.2bn while goods spending fell $49.9bn. The personal saving rate rose to 3.0% from 2.6% in June — the largest monthly gain this year, from a level that has been sitting near multi-decade lows. Income arrived, durable goods purchases were deferred, and the difference was saved.

That combination — an inflation gauge stalled above target and a real consumption line at zero — is the most consequential thing in the release, and it is not a currency story that runs through one factor. It runs through two of the five in opposite directions: the interest-rate factor reads a rate that has not eased, and the growth factor reads a consumer that stopped.

Goods, services and where the pressure actually sits

Beneath the aggregate, the composition is not the one a tariff story would predict.

Goods prices fell on the month. Housing rose 0.27% and health care 0.20%, both unremarkable. The category doing the work is services excluding energy and housing, at 0.28% on the month and 3.85% on the year — and, as the previous section established, a meaningful share of that is the financial line rather than a broad acceleration in what service providers charge. "Other services" — a large, genuinely market-priced grouping — rose 0.106% and runs at 2.39% annually, which is below the Federal Reserve's target on a twelve-month basis.

The 30 September rewrite, now one print away

The preview flagged that the number had a known expiry date. Today's release tightened the timetable in a way worth noting: BEA confirmed that the next Personal Income and Outlays report, covering August, publishes on 30 September 2026 — the same day the 2026 annual update of the national accounts begins.

That update revises the statistics from the first quarter of 2021 through the first quarter of 2026, and changes three consumer deflators:

So July's 3.344% is the last core reading produced entirely on the current method, and the August print will arrive already restated. Moving portfolio management from an asset-value basis to an employment-based extrapolator severs precisely the link this article has been describing — a history in which core PCE was partly an equity index becomes a history in which it was not. What that does not change is the labour, shelter and core-goods data, which revise very little and are what a central bank actually reacts to.

Which scenario landed, and what it does to the dollar

The preview set out three cases. The hot one landed — monthly core at 0.246%, annual rate holding at 3.3% — but with the composition caveat that made the hot case ambiguous, and with a spending number nobody had in the scenario map.

The July FOMC held the target range at 3.50–3.75% by a 9–3 vote on 29 July, the fifth consecutive meeting without a change, with three dissents preferring a quarter-point hike and no forward guidance in the statement. Going into this print the front end was already carrying a tightening bias that the policy rate does not justify: on the Treasury's own daily par yield curve, the two-year note yielded 4.17% on 25 August against a 3.625% target midpoint, with the ten-year at 4.64% and the 30-year at 5.17% after the move traced in the long-end selloff.

The honest reading is that this report did not resolve September; it sharpened the disagreement and added a growth question to it. Which is why the next scheduled event matters more than it did yesterday: Chair Warsh's Jackson Hole keynote on 28 August is now the first opportunity to hear which of these two readings the Committee is taking — a stalled headline rate, or a stalled consumer paying an inflation rate that is mostly measured somewhere other than the checkout.

Bottom line

The release was summarised in one number, and that number was partly an equity index wearing a trench coat — verifiably so, at about 0.12 percentage points of a 0.246-point month. The durable content sits in the categories that were not the financial line: goods prices negative, housing at 0.27%, health care at 0.20%, other services at 2.39% annually. Set against a real consumption line at zero and a saving rate rising into it, the report describes an economy where the measured price level is stalling above target for reasons that partly are not about consumer prices, while the actual consumer has stopped. Everything the dollar page tracks through the rate channel now depends on which of those two facts the Committee weights — and, as always, the useful question is not what will print, but what would have to be true for the print to change a policy path. The June instalment of this series is here, the July CPI breakdown that fed into it is here, and more on the method behind that framing is on the about page.

Educational macro context only — not investment advice.