Core CPI 0.3% Against 0.2% Expected (11 September 2026): August Inflation Firmed on the Month and Eased on the Year — Four Days Before the Fed Decides

August's Consumer Price Index landed at 8:30 a.m. Eastern on Friday 11 September with the headline exactly where the market expected it — 0.4% on the month, 3.4% on the year — and the core line a tenth above it, at 0.29% against a 0.2% consensus. The annual core rate nonetheless fell, from 2.5% to 2.4%, which is the number most readers will see. Both facts are true and they point in opposite directions, because one describes the last twelve months and the other describes the last four weeks. The monthly figure is the one the Federal Open Market Committee will read on Tuesday, and it broke a three-month run of soft core prints in the week the Committee was already a two-thirds bet to tighten.

What actually happened

Sources: actuals calculated from BLS index levels retrieved via the BLS public data API (all items, seasonally adjusted, 334.131 in August against 332.813 in July; core 337.765 against 336.789). Consensus per the Dow Jones survey as reported by CNBC, whose headline read that core inflation was higher than estimated. Nowcast from the Cleveland Fed, updated 8 September.

Three of the four cells came in as forecast. The fourth is the release. A tenth of a percentage point on a monthly core reading sounds like nothing, and in most months it is; in this one it is the difference between a Committee that has watched core inflation decelerate for three consecutive months and a Committee that has not.

The tenth the model missed, and where it came from

The interesting question is not that core beat by a tenth but which part of core did it, because the candidates carry completely different implications.

It was not shelter. Rent of primary residence rose 0.174% on the month and owners' equivalent rent 0.186% — both soft, both consistent with the year-long deceleration that has taken annual shelter inflation from above 4% to 3.04%, down again from 3.17% in July. The single largest disinflationary force in the American price data did not stop working in August.

It was transport and goods.

Calculated from BLS index levels via the public data API.

Airline fares are the cleanest energy-to-core channel the CPI contains: jet fuel is a large, visible share of an airline's operating cost, and the fare line has now risen 23.4% over twelve months while the crude benchmark it tracks pushed through $100. Core goods, the channel most exposed to freight, had been the standing evidence that pass-through was not happening — over the three months to July it ran at an annualised −0.01%. Over the three months to August it runs at +0.87%.

That is the whole story of this print. For six months the argument that the energy shock was contained rested on core goods going nowhere and core services ex-shelter behaving. In August, the first of those two stopped being true.

The 2026 episode, in one table

Source: BLS Consumer Price Index, seasonally adjusted series, via the BLS public data API. The published unadjusted headline for August was 3.4%.

On the annual view nothing happened in August. Core sits at 2.45%, still below January's 2.51%, still well inside the range it has occupied all year. A reader who checks this table once a month would conclude the energy shock never touched underlying inflation and has now passed.

That reading was correct through July and it is the one this release complicates. The annual series is smooth because it is an average; the month underneath it is not. Energy rose 2.10% on the month and 16.05% over the year, with gasoline +3.9% and +27.4% — and for the first time this cycle, the lines that sit downstream of the barrel moved with it.

The gauge this release doesn't publish

The Federal Reserve's 2% objective is not defined on the Consumer Price Index. It is defined on the Personal Consumption Expenditures price index, produced by the Bureau of Economic Analysis, and the two measures have been telling different stories all year.

Sources: BLS Consumer Price Index, July 2026; BEA Personal Income and Outlays, July 2026, released 26 August.

That the Fed's own gauge sits above the widely quoted one is the reverse of the usual arrangement, and three structural differences explain it. Weights: shelter is close to half of core CPI and roughly half that share in the PCE, so CPI's decelerating rent line does twice the work in the CPI. Scope: the CPI measures what urban consumers pay out of pocket, while the PCE includes the large share of health care paid on households' behalf by employers and government programmes. Sources: for hospital care and physician services, the BEA takes prices from the Producer Price Index rather than the CPI.

August widened the gap from both ends simultaneously, and this is the part worth sitting with. Friday's CPI showed medical care services falling 0.248% on the month. Thursday's PPI showed the hospital line the BEA actually uses — general medical and surgical hospitals — accelerating to 3.68% over twelve months from 3.46%, with offices of physicians at 1.21%. The CPI's medical reading and the PCE's medical reading moved in opposite directions in the same month, because they are different surveys measuring different transactions.

So the Committee arrives on Tuesday holding a core CPI that beat by a tenth and a core PCE whose distinctive inputs firmed independently. Thursday's producer report, whose 5.4% annual headline the wires led with, mattered for this reason and not for that number: final demand goods rose 1.1% against services at 0.1%, and core PPI excluding food and energy rose 0.2%, a tenth below estimate. The barrel did the work in the headline. The hospital lines did the work in the gauge that counts.

Gasoline was the half you could compute

Worth recording, because it demonstrates how little information a headline CPI print carries. The Energy Information Administration publishes a national average retail gasoline price every week, and by 1 September all five of August's weeks were public: an August average of $4.192 against July's $4.064, up 3.1% before seasonal adjustment.

That arithmetic, set out here before the release, implied gasoline's annual rate would rise from 24.6% towards roughly 28%. It printed 27.4%, with the seasonally adjusted monthly change at +3.9%. The energy contribution to Friday's headline was knowable a week in advance from a free weekly series.

This is the practical lesson of the release. The 3.4% headline that leads the coverage was largely public information on 1 September. The 0.29% core line, which nobody could compute in advance, is where the entire information content sat — and it is the line that has to be read to one more decimal place than the wires print it to, because 0.29% and 0.25% round to the same number and mean different things.

What the print does to Tuesday

Set out in advance here were three branches. The one that landed was the third: core at or above +0.3%, described then as "the tail that would actually move things." It is worth being precise about why, and about the limits.

The mechanism runs through the front end of the Treasury curve and nothing else. Inflation does not lift a currency; a repriced policy path does. Going into the release the market had roughly a 66% probability of a quarter-point hike on 16 September priced, up from about 56% on Wednesday, after Thursday's PPI and $100 crude took the two-year note up 13 basis points to 4.56% and the ten-year to 4.95%. The two-year is the leg that transmits to the dollar, because carry is earned on short-dated rates.

Two constraints on how far that can run, both of which cut against the obvious reading:

Most of the distance was already travelled. A market that has moved from 56% to 66% in a session has spent the cheap part of the repricing. The remaining move to a fully priced hike is smaller than the one already made, which mechanically limits how much a confirming data point can add. This is the asymmetry noted here before the print, and Friday resolved it in the direction that had the least room.

The Committee has told the market it looks through this channel. The July statement attributed above-target inflation to supply shocks in certain sectors, including energy. The core beat was driven by airline fares and core goods — which is to say, by energy arriving one step downstream. A Committee that meant what it wrote in July has a ready-made reason to discount exactly the component that firmed. Whether it does is the question Tuesday answers, and nobody can ask it in the meantime: the blackout that began on 5 September runs through 17 September.

What is not in doubt is that the strongest single argument against tightening — a three-month annualised core rate of 1.64% — is no longer available in that form. It is 1.97% now.

What would change the picture

The dot plot, not the decision. The 16 September projections carry more information about 2027 than any single print does about 2026. A hike delivered alongside an unchanged median for next year is a different signal from the same hike with the path lifted.

Whether core goods was one month or a trend. August took the three-month annualised core goods rate from −0.01% to 0.87% on a single 0.106% monthly print. That is a small number doing a lot of work in an annualised calculation, and one month is not a trend. September's report, due 14 October, is the confirmation or the correction.

Shelter. Still decelerating at 3.04% and still the reason core CPI reads below core PCE. A stall here would close the gap from the wrong end, and would matter more than anything airline fares do.

The September energy base. The same EIA weekly series that made August computable is already building October's report. The Cleveland Fed's September nowcast sits near 0.40% headline against 0.19% core — another month in which the two lines are set to say opposite things.

The labour-market side of this decision, where the August payroll beat of 162,000 removed the argument for waiting, is covered here; the inflation test the Chair set at Jackson Hole is set out here. The euro side of the same energy shock, with the ECB having decided on 10 September, is in the ECB note, and the fuller account of why the Fed's preferred gauge and the CPI keep diverging is in the PCE note.

For the currency the transmission is the interest-rate factor and nothing else. The factor-by-factor read on the dollar sits on the USD page, with the euro's on EUR and the yen's on JPY — the two crosses that have carried most of the dollar's 2026 range. What Pip Theory is and how it reads a release like this are set out on the about page.

The headline said 3.4% and was public in advance. The annual core said 2.4% and fell. The month underneath said 0.29% and rose. All three are the same release, and only the last one was news.

Educational macro context only — not investment advice.