Economics · Economic indicators
The CPI report: how BLS builds it and how to read it
How the Consumer Price Index is built (basket, weights, shelter and owners’ equivalent rent), what headline and core mean, and how to read m/m and y/y changes.
On this page
The Consumer Price Index (CPI) report is the most closely watched inflation release in the US. Once a month, at 8:30 a.m. Eastern time, the Bureau of Labor Statistics (BLS) says how much the prices urban consumers pay changed over the past month and the past year. The September 2026 report, released on October 14, 2026, showed consumer prices up 0.6% in the month and 3.7% over twelve months, with core prices, which exclude food and energy, up 0.3% and 2.6%.
Those few numbers move bond yields, the dollar and stock index futures within seconds. They also set the principal of inflation-protected Treasuries and feed into wage contracts and rents. Yet most people who react to the headline never look at how it is built: what goes in the basket, how each item is weighted, and why housing, the largest piece, is measured from rents rather than home prices.
This page walks through that machinery and shows how to read a release, using the September 2026 report as the worked example. For the bigger picture of what inflation is and why it matters, start with Inflation: what it is, how it’s measured and what it does to money.
What the CPI measures
The CPI is an average of the prices paid for a fixed list of goods and services, weighted by how much consumers spend on each. BLS publishes several versions:
- CPI-U (all urban consumers) is the headline index. Its population is urban households, which BLS says make up over 90% of the US population. It leaves out rural nonmetropolitan residents, farm households, people on military bases and people in institutions such as prisons.
- CPI-W (urban wage earners and clerical workers) covers a subset of about 30% of the population: households where most income comes from hourly-type occupations. It excludes professionals, the self-employed, the unemployed and retirees.
- C-CPI-U (chained CPI-U) uses a formula that tracks consumers switching between items more closely. Unlike the other two, it is revised: BLS finalizes it 10 to 12 months after first publication.
The CPI prices consumption, not investment or taxes. Income and Social Security taxes, stocks, bonds, real estate purchases and life insurance are outside its scope. That last point is why a house purchase does not enter the CPI directly, which matters a great deal for shelter (below).
How BLS builds the CPI-U
Building the index takes three ingredients.
- A sample of prices. BLS data collectors and online sources price specific items, in specific stores and areas, through the month. Collection is spread across three pricing periods of roughly ten days each, and BLS records about 80,000 items a month. The same items are re-priced over time so that the index measures price change, not a change in what was bought.
- Weights. Each of more than 200 item categories gets a weight equal to its share of consumer spending, from the Consumer Expenditure Surveys run for BLS by the Census Bureau. In the 2021 survey year more than 20,000 households answered quarterly interviews and about 12,000 more kept two-week spending diaries. Since the January 2023 index, BLS updates the weights every year, using spending from two years earlier; before that it updated them every two years.
- Aggregation. Price changes within each category are combined, then the categories are rolled up with their weights into eight major groups and the all-items index.
The weights drift during the year as relative prices change: a category whose prices rise faster gains share. BLS publishes these shares each December as the "relative importance" table.
What is in the basket
The main components of the CPI-U, from BLS’s December 2025 relative-importance table (2024 weights):
| Component | Share of all items, % |
|---|---|
| Services less energy services | 60.7 |
| — Shelter | 35.6 |
| —— Owners’ equivalent rent of residences | 26.2 |
| —— Rent of primary residence | 7.8 |
| — Medical care services | 6.9 |
| — Transportation services | 6.3 |
| Commodities less food and energy | 19.2 |
| — New vehicles | 3.8 |
| — Used cars and trucks | 2.8 |
| — Apparel | 2.4 |
| Food | 13.7 |
| Energy | 6.4 |
| — Gasoline (all types) | 2.9 |
| All items less food and energy (core) | 79.9 |
Two facts jump out. Services make up most of the index, and shelter alone is more than a third of it. A 1-point change in shelter inflation moves headline CPI by about 0.36 point; a 1-point change in apparel inflation moves it by 0.02 point.
Shelter and owners’ equivalent rent
Many US households own their home, so the CPI needs a way to price the housing they consume. BLS treats a house as a capital good, an investment, separate from the shelter service it provides. House prices, mortgage interest, property taxes and most maintenance are therefore not in the CPI. Instead, BLS estimates owners’ equivalent rent (OER): what owners would pay to rent their own home, unfurnished and without utilities.
The mechanics, from BLS’s rent fact sheet:
- The weight of OER comes from a Consumer Expenditure Survey question asking owners what their home would rent for.
- The price change comes only from rental units. Owner-occupied homes are not priced.
- Each sampled rental unit is priced every six months. The sample is split into six panels, each priced in two months of the year (panel 1 in January and July, for example).
That design makes shelter inflation slow and smooth. Most tenants’ rents change once a year at lease renewal, and each unit is visited twice a year, so a turn in market rents can take many months to show up fully in the CPI. Analysts who want an earlier read look at private new-lease rent indexes, but those are not part of the official measure.
Headline and core
The headline CPI is the all-items index. Core CPI removes food and energy, whose prices are the most volatile, to show the underlying trend. The chart shows why both are watched. In 2021 and 2022 headline ran well above core as energy prices surged, then fell below it as energy prices dropped in 2023. In 2026 the gap opened again: by BLS’s figures, headline was 3.7% in September while core was 2.6%. Core inflation gets its own page: Core inflation: why food and energy are left out.
Reading the September 2026 report
The release’s summary table gives one-month (seasonally adjusted) and 12-month (unadjusted) changes for each component. For September 2026, BLS reported:
| Component | 1-month, SA | 12-month, NSA |
|---|---|---|
| All items (headline) | +0.6% | +3.7% |
| Core (ex food/energy) | +0.3% | +2.6% |
| Energy | — | +20.4% |
| Food | — | +2.5% |
| Shelter | — | +3.2% |
BLS also noted that gasoline accounted for over half of the monthly all-items increase. (The dashes are figures not quoted on this page, not missing data.)
A worked example: contributions to the headline
A component’s contribution to headline inflation is roughly its weight times its own inflation rate. Using the December 2025 weights and the September 2026 12-month changes:
| Component | Weight | 12-month change | Contribution, points |
|---|---|---|---|
| Core | 79.9% | 2.6% | 2.08 |
| Energy | 6.4% | 20.4% | 1.30 |
| Food | 13.7% | 2.5% | 0.34 |
| Total | 3.72 |
The total lands at about 3.7%, matching the reported headline. The table makes the story plain: energy is only 6.4% of the basket but contributed about 1.3 of the 3.7 points, more than a third of headline inflation. Shelter, inside core, contributed about 1.1 points (35.6% × 3.2%). The method is an approximation, because the weights drift through the year, but it is a fast way to see what is driving a print.
Month over month or year over year?
The y/y rate compares this month with the same month a year ago. It is smooth and is the number most headlines quote, but it is slow: eleven of its twelve months are old news. The m/m rate is current but noisy. To compare a monthly change with an annual goal, compound it:
A single month annualized is a dramatic number and usually a misleading one; one hot or cold month says little. Many analysts prefer 3-month or 6-month annualized rates as a middle ground. Two more traps:
- Rounding. BLS publishes one decimal. A "0.3%" core month can be anything from 0.25% to 0.349%, which annualizes to anywhere between 3.0% and 4.3%.
- Base effects. The 12-month rate changes both when a new month comes in and when an old month drops out. A big jump a year ago falling out can pull y/y inflation down even if the latest month was normal.
The missing October 2025
A lapse in federal appropriations ran from October 1 to November 12, 2025, and BLS could not collect October 2025 prices. For most CPI series, including the all-items and core indexes, no October 2025 value was published. For the November indexes, BLS carried October prices forward from September; rents and OER were held unchanged for October. BLS notes that carrying prices forward may understate inflation when prices are generally rising.
Two practical consequences follow. First, BLS has said it will not publish 12-month changes for October 2026 for indexes where October 2025 data are missing. Second, anyone computing 12-month changes from a downloaded table must match calendar months. Counting back twelve rows skips the gap and silently produces 13-month changes for every month from November 2025 to September 2026.
Pitfalls and limitations
- It is an average. Your own inflation depends on your spending. A renter, a homeowner with a fixed-rate mortgage and a commuter facing $4 gasoline see different numbers.
- Shelter lags. Because of OER’s design, shelter inflation reflects rents set over the past year or more, not today’s market.
- Quality adjustment. BLS adjusts some prices for quality changes. The adjustments are necessary and imperfect, and the subject of persistent debate.
- Seasonal adjustment. Monthly changes are seasonally adjusted; 12-month changes in the release use unadjusted data. Mixing them up causes small but confusing differences.
- CPI is not the Fed’s target. The Fed’s 2% goal is defined on the PCE price index, which uses broader scope, different weights and a different formula. See PCE inflation: the Fed’s preferred gauge vs the CPI.
CPI versions and neighbours compared
| Measure | Who | Population or scope | Revised? |
|---|---|---|---|
| CPI-U | BLS | Urban consumers, over 90% of population | No (unadjusted data are final) |
| CPI-W | BLS | Wage-earner households, about 30% | No (unadjusted data are final) |
| C-CPI-U | BLS | Same as CPI-U, chained formula | Yes, final after 10–12 months |
| PCE price index | BEA | Spending by and on behalf of households | Yes |
| PPI final demand | BLS | Prices received by domestic producers | Yes |
How Finabase does it
Finabase does not carry CPI data, component tables or release dates, and it has no economic calendar. The two charts on this page are ours, built from public BLS series downloaded via FRED on October 30, 2026.
What the app shows is the market reaction. The Treasury page lists yields from 1 month to 30 years, and the 2-year yield, which closely tracks expected Fed policy, often moves most on an inflation surprise. The 10-year Treasury yield page has its own chart. Commodity futures behind the energy component, such as WTI crude oil, are on the futures page. Coverage of each release appears in the economy topic of Discover.
Frequently asked questions
What time does the CPI report come out?
BLS releases it at 8:30 a.m. Eastern time. The September 2026 report came out on October 14, 2026, and BLS scheduled the October 2026 report for November 10, 2026. The full calendar is on bls.gov.
Does the CPI include house prices?
No. BLS treats a home as an investment, so purchase prices and mortgage interest are out of scope. Homeowners’ housing costs enter as owners’ equivalent rent, an estimate of what their home would rent for, which was 26.2% of the CPI-U in December 2025.
Why is shelter such a big part of the CPI?
Because housing is the largest item in most household budgets. Shelter, which combines OER, rent and smaller items such as hotel stays, was 35.6% of the CPI-U in December 2025.
What is the difference between CPI-U and CPI-W?
CPI-U covers all urban consumers, over 90% of the population. CPI-W covers a narrower group of wage-earner households, about 30% of the population, so its weights differ slightly.
Is the CPI revised?
The unadjusted CPI-U and CPI-W are final when issued. The chained C-CPI-U is revised and finalized 10 to 12 months later.
Summary
The CPI report measures the prices urban households pay, using about 80,000 monthly price quotes weighted by spending shares that BLS updates every year. Shelter, mostly owners’ equivalent rent estimated from rental units priced twice a year, is about 36% of the index, so housing dominates the slow-moving core while energy, only about 6%, dominates the monthly swings. In September 2026 headline CPI rose 3.7% over the year, with energy up 20.4% contributing about a third of it, and core rose 2.6%. Read the monthly figure for news, the 12-month figure for the trend, and annualized 3- or 6-month rates in between, and remember that the October 2025 data gap means no 12-month change for October 2026 in most indexes.
Sources and further reading
- BLS — Consumer Price Index news release, September 2026 (released October 14, 2026)
- BLS — Consumer Price Index frequently asked questions
- BLS — Relative importance of components in the CPI, December 2025
- BLS — Relative importance and weight information for the CPIs
- BLS — Measuring price change in the CPI: Rent and rental equivalence
- BLS — 2025 federal government shutdown impact on the CPI (FAQ)
- BLS — CPI-U all items, SA (FRED series CPIAUCSL), downloaded 2026-10-30
- BLS — CPI-U less food and energy, SA (FRED series CPILFESL), downloaded 2026-10-30
- BLS — CPI-U energy, shelter and food, SA (FRED series CPIENGSL, CUSR0000SAH1, CPIUFDSL), downloaded 2026-10-30