How the figures are built

Every index on this site is a value published by the Bureau of Labor Statistics, read from its files and never typed by hand. This page lists which files, which series, how a missing month is handled, the formulas, and the tests that run before each release of the site. Data file of October 19, 2026, indexes through September 2026.

The data

The indexes come from the BLS time series files for the CPI (download.bls.gov/pub/time.series/cu, plus the cw files for the CPI-W and the su files for the chained CPI-U). The script scripts/data/build-cpi.py reads them and writes a frozen file, dated, that the pages and calculators load. A refresh can also go through the BLS Public Data API version 2, whose terms ask users to cite the date of retrieval and to state that BLS.gov cannot vouch for the data or analyses derived from these data after the data have been retrieved from BLS.gov. We do both.

We keep 30 areas: the U.S. city average (monthly since 1913), the 4 census regions (since 1966), the 2 size classes (Size A, Size B/C) and the 23 metropolitan areas, each with its own start date. For each area we keep the all-items index and 12 more, from 1997: the 8 major groups of the CPI, shelter, food at home, energy and all items less food and energy. All are CPI-U, not seasonally adjusted, which is the version used for escalation clauses and the one the BLS never revises.

The area code replaces : 0000 for the U.S., S12A for New York, and so on. Each area page names its code.
Index	BLS series
All items	`CUUR<area>SA0`
Housing	`CUUR<area>SAH`
Shelter	`CUUR<area>SAH1`
Food and beverages	`CUUR<area>SAF`
Food at home	`CUUR<area>SAF11`
Energy	`CUUR<area>SA0E`
Transportation	`CUUR<area>SAT`
Medical care	`CUUR<area>SAM`
Apparel	`CUUR<area>SAA`
Recreation	`CUUR<area>SAR`
Education and communication	`CUUR<area>SAE`
Other goods and services	`CUUR<area>SAG`
All items less food and energy	`CUUR<area>SA0L1E`

Alongside: the CPI-W, U.S. city average (CWUR0000SA0), used for Social Security, and the chained CPI-U (SUUR0000SA0, base December 1999=100), whose last 10 to 12 months are preliminary and replaced at each refresh.

Months that do not exist

Two kinds of month have no index, and the site fills neither.

Months an area is not priced. Only New York, Chicago, Los Angeles are published every month. Boston, Minneapolis-St. Paul, Washington, DC, Tampa, Dallas-Fort Worth, Denver, Riverside, San Diego, Urban Hawaii are published in odd months; Philadelphia, Detroit, St. Louis, Miami, Atlanta, Baltimore, Houston, Phoenix, San Francisco, Seattle, Urban Alaska in even months. When you choose such a month, the calculator keeps it in the list with a label, uses the latest earlier month published and says so. Example: June 2024 in Boston uses May 2024.

October 2025. No prices were collected during the lapse in appropriations (October 1, 2025 to November 12, 2025), and the BLS publishes a dash with the footnote "Data unavailable due to the 2025 lapse in appropriations" for every series. The calculator uses September 2025 and shows the November index beside it. The 12-month change for October 2026 will not be computed either, because its base is missing (BLS questions and answers). An annual average for 2025 is the one the BLS published, from 11 months, never one we recompute.

Dates past the latest release or before the first index. The calculator uses the latest month (September 2026) or the first month of the series, and says so. It never extrapolates.

Formulas

The value of an amount A between two dates is A × CPI(to) ÷ CPI(from), with the full published index, to three decimals, rounded only at display. Example: $1,000 of June 2015 in the U.S. index (238.638) is $1,411.93 in September 2026 (336.94), a change of 41.2 %. The average annual rate, shown when the span is a year or more, is (CPI(to) ÷ CPI(from)) to the power 12 ÷ months, minus 1. The 12-month change of a month compares it with the same month a year earlier, as the BLS does; for a bimonthly area it uses the same month of the previous year, which is published on the same rhythm. Year-on-year inflation for a past year compares the two annual averages published by the BLS.

The Social Security COLA follows 42 U.S.C. 415(i): the average CPI-W of July, August and September, rounded to three decimals, compared with the same average in the last year a COLA was paid, rounded to the nearest tenth of a percent. For 2026: 328.746 against 317.265, which gives 3.6 %, the figure the SSA announced on October 14, 2026 (SSA press release).

The checks

The test suite (npx vitest run) must pass before every release of the site. It checks, among others: that the all-items index of September 2026 and the CPI-W of the same month equal the values of the BLS release of October 14, 2026; that the 12-month changes computed for all items, core, energy and food at home match the release; that each of the 30 areas has all 13 indexes; that 3 metros come out monthly and 20 bimonthly; that October 2025 is absent from every series and never filled; that the 2025 annual inflation recomputed from the published averages is 2.6 %; and that the COLA formula rebuilds the SSA table from 1984, when the third-quarter comparison began, to 2026 (earlier COLAs compared first quarters). It does, except for 1990 and 1998 where the published figures (5.4 % and 1.3 %) differ by a tenth from the formula on the final index, and 1999, which a law set at 2.5 %. Content tests check every page: length of titles and descriptions, the opening answer, the FAQ, and the absence of unknown values.

Limits

A local CPI measures price change in one place over time; it does not compare price levels between places. Indexes for metros come from smaller samples than the national one and swing more from one release to the next; the BLS advises caution with short periods. Averages are weighted by the spending of all urban households and differ from any one household's experience. Amounts in the results are rounded to the cent and are estimates, not advice.

Sources

Results apply the published CPI indexes to the amount you type. An index tracks the average prices paid by urban consumers in an area, not your own budget, and it does not compare price levels between cities. A month the BLS did not publish is never filled in: the calculator names the index it uses instead.