How to read the US CPI report, using the September 2026 figures

When it comes out

BLS publishes the CPI at 8:30 a.m. Eastern time, with the release embargoed until that minute. The September 2026 report (release number USDL-26-1574) came out on Wednesday 14 October 2026. BLS’s CPI release schedule lists the next ones as 10 November 2026 for October data, 10 December 2026 for November data and 13 January 2027 for December data, all at 8:30 a.m. The same dates appear in the BLS calendar feed our site reads, last checked on 19 October 2026, and in the “Today” strip of official schedules at the top of each page.

One gap matters for anyone redoing the arithmetic next month. In the BLS data series, the October 2025 index value is blank, footnoted as unavailable because of the 2025 lapse in appropriations. The usual 12-month comparison for October 2026 needs that value, so it cannot be recomputed from published index levels the way we do below. The 10 November release will show how BLS presents that month.

What the release contains

The first page gives the all items CPI-U change for the month (seasonally adjusted) and for the last 12 months (not seasonally adjusted), then the main drivers: food, energy and the index excluding both, often called core. Table A sets out seasonally adjusted monthly changes for the last seven months for each major group, with the unadjusted 12-month change in the final column. Detailed tables follow, with index levels to three decimal places.

The release covers more than one index. CPI-U, for all urban consumers, reflects the spending of a population BLS puts at over 90% of the US total. CPI-W, for urban wage earners and clerical workers, covers a subset of about 30% of the population, defined by where household income comes from. CPI-W rose 3.9% over the 12 months to September 2026, to 330.654, against 3.7% for CPI-U, and one visible reason is that energy carries a larger weight in CPI-W (8.052% against 6.383% in BLS’s December 2025 relative importance table). The chained CPI-U, a third measure, rose 3.5% over the year and is issued in preliminary form, subject to revision.

Seasonal adjustment, revisions, and why the 12-month rate is unadjusted

Some prices move in a pattern that repeats every year. BLS lists weather, production cycles, model changeovers, holidays and sales as typical causes. Seasonal adjustment removes that typical pattern so a monthly change shows what was unusual for the time of year. BLS computes seasonal factors with the X-13ARIMA-SEATS method and, for some series such as motor fuels, first strips out extreme moves (intervention analysis) so they do not distort the factors.

The factors are updated each February, and the new factors are used to revise the previous five years of seasonally adjusted data. In the January 2026 release, BLS recalculated seasonally adjusted indexes for 2021 to 2025. A seasonally adjusted monthly change you read today can therefore be restated later; the unadjusted CPI-U and CPI-W are final when released.

That is why the 12-month rate uses unadjusted data. Over a full year the seasonal pattern largely cancels out, and unadjusted indexes are not revised, and BLS notes that many wage contracts and pension plans tie their adjustments to unadjusted CPI; it advises against seasonally adjusted data in escalation agreements because those series are revised. In September 2026 the two give the same rounded answer: the seasonally adjusted 12-month change works out to 3.717% (336.298 ÷ 324.245 − 1, our calculation) and the unadjusted to 3.738%, both 3.7%.

A worked example: September 2026 from the index levels

BLS works in percent changes, not index points, because a point move means different things at different index levels. The rule is new index ÷ old index − 1, then rounded to one decimal place. All index levels below are from the BLS public data API, base 1982–84 = 100.

Headline, monthly, seasonally adjusted (series CUSR0000SA0): 336.298 ÷ 334.131 − 1 = 0.649%, published as 0.6%. August was 334.131 ÷ 332.813 − 1 = 0.396%, published as 0.4%.

Headline, 12 months, not seasonally adjusted (CUUR0000SA0): 336.940 ÷ 324.800 − 1 = 3.738%, published as 3.7%. The August figure was 334.980 ÷ 323.976 − 1 = 3.397%, published as 3.4%.

Core, monthly, seasonally adjusted (CUSR0000SA0L1E): 338.896 ÷ 337.765 − 1 = 0.335%, published as 0.3%. Core, 12 months, not seasonally adjusted (CUUR0000SA0L1E): 339.287 ÷ 330.804 − 1 = 2.564%, published as 2.6%, up from 2.4% in August.

Rounding hides detail. September’s headline monthly change of 0.649% sits just under the 0.65% line, and a figure of 0.551% would also print as 0.6%. The unrounded figure, which anyone can compute from the index levels, is a fuller description of the month. BLS also publishes a standard error for the monthly change: it gives 0.04 percentage points for the all items index, so a 95% confidence band is about plus or minus 0.08.

Which components moved the total

A component’s effect on the total depends on its change and on its weight, which BLS publishes as relative importance. In the December 2025 table (2024 weights), shelter was 35.625% of CPI-U, food 13.698% and energy 6.383%. Relative importance drifts each month as relative prices change, so we updated each weight to August 2026 with the unadjusted indexes: weight × (component index Aug 2026 ÷ Dec 2025) ÷ (all items index Aug 2026 ÷ Dec 2025).

For energy that is 6.383 × (329.351 ÷ 274.131) ÷ (334.980 ÷ 324.054) = 7.42%. Multiplying by September’s seasonally adjusted change of 5.196% gives about 0.39 percentage points, roughly 60% of the 0.649% total. The same method gives shelter about 0.14 points (35.32% × 0.391%) and food under 0.01 points (13.51% × 0.048%). These are our approximations: they mix unadjusted weights with seasonally adjusted changes, and BLS does not publish contributions in the release. They agree with the release’s own account, which says gasoline rose 9.1% in September, “accounting for over half of the monthly all items increase.”

Energy also explains the gap between headline and core over the year. The energy index was up 20.4% over the 12 months to September, with gasoline up 34.2%, while core rose 2.6%.

CPI and the Fed’s PCE measure

The Federal Open Market Committee’s Statement on Longer-Run Goals defines its 2% inflation goal in terms of the annual change in the price index for personal consumption expenditures (PCE), which the Bureau of Economic Analysis (BEA) publishes in its Personal Income and Outlays release. PCE is a separate index with its own coverage and weights, so its rate and the CPI rate for the same month need not match, and a CPI release does not settle what PCE will show.

BEA’s release calendar schedules Personal Income and Outlays for September 2026 on 29 October 2026 at 8:30 a.m. ET (12:30 UTC in the feed). Our central bank tracker at /central-banks shows the Fed’s target range and next meeting, and /rates shows Treasury yields and the effective fed funds rate with their dates.

How to check the numbers yourself

Each figure above can be rebuilt from public pages.

• The release: bls.gov/news.release/cpi.nr0.htm (latest month), with Table A on the first page.
• Index levels: the BLS data API or data.bls.gov, series CUSR0000SA0 and CUUR0000SA0 for all items, CUSR0000SA0L1E and CUUR0000SA0L1E for core, CWUR0000SA0 for CPI-W. FRED mirrors the seasonally adjusted series as CPIAUCSL and CPILFESL, which matched BLS on 21 October 2026.
• Weights: the BLS relative importance page, December 2025 table.
• Dates: the BLS CPI release schedule page or its calendar feed, and BEA’s release schedule for PCE.
• Redo the arithmetic as new index ÷ old index − 1, rounded to one decimal, and compare it with what BLS printed.