The CPI Report: What Gets Revised and What Never Changes

Inflation data is usually described to traders as a number that arrives, moves the market, and then belongs to history. Price data does not date the economic cycle directly, but it is what converts nominal income and sales into the real measures that do.

One half of the consumer price index works exactly like that and never changes again. The other half, which is the half quoted in almost every report of the monthly figure, keeps being rewritten for five years.

Which is which is documented, and it is the opposite way round from what most explanations imply.

What the CPI Report Measures, and Who It Measures It For

The consumer price index measures the change in prices paid by consumers for goods and services. The word change carries the weight: it is an index of movement from a base period, not a price level.

It also measures that change for a defined population rather than for everyone, and there are two of them.

The all urban consumer group represents over 90 per cent of the total US population. It covers almost all residents of urban or metropolitan areas, including professionals, the self-employed, the poor, the unemployed and retired people.

The CPI-W is narrower. It covers households inside that definition where more than half of income comes from clerical or wage occupations and at least one earner was employed for at least 37 weeks during the previous 12 months, and it represents approximately 30 per cent of the population.

Some groups are in neither. BLS states that the spending patterns of people in rural nonmetropolitan areas, farming families, people in the Armed Forces and those in institutions such as prisons and mental hospitals are not included.

A third index, the Chained CPI for All Urban Consumers, is published at national level only and behaves differently again on revisions, which the later section covers.

The Sample Behind the Number

Descriptions of the index rarely say where the prices come from, which makes the figure sound more like a measurement of everything than it is.

Prices are collected each month in 75 urban areas across the country, from about 6,000 housing units and approximately 22,000 retail establishments including department stores, supermarkets, hospitals and other service establishments.

Collection is not uniform across the country either. Fuels and a few other items are priced every month in all 75 locations, while most other commodities and services are collected every month in the three largest geographic areas and every other month elsewhere.

Prices are obtained by personal visit, telephone call, web or app collection by trained BLS representatives, and all taxes directly associated with the purchase and use of items are included in the index.

Price changes for the items in each location are then aggregated using weights representing their importance in the spending of the relevant population group, and local data are combined into a US city average.

What Is Final on Release and What Is Not

This is the part that separates the CPI from most other headline releases, and it runs in a direction that surprises people who follow the labour market data.

BLS states that the CPI-U and CPI-W are considered final when released. The unadjusted index level published for a month is the number, and it stays the number.

The Chained CPI does not share that property. It is issued in preliminary form and is subject to three subsequent quarterly revisions.

And the seasonally adjusted series, which is what a monthly change is normally quoted from, is revised for up to five years after its original release.

The comparison with the labour market release is instructive rather than academic. Payrolls arrive preliminary and settle later, as our guide to the payrolls report sets out, while the unadjusted CPI arrives settled and the adjusted version is what moves.

Anyone building a record of past releases needs to know which of the three they stored, because two of them may no longer match what BLS currently publishes for the same month.

Why the Seasonally Adjusted Series Keeps Changing

Seasonal adjustment removes effects that occur at about the same time and in about the same size every year, such as weather events, production cycles, model changeovers, holidays and sales.

The factors that do this are estimated, and BLS recalculates them. Seasonally adjusted data are computed using factors derived by the X-13ARIMA-SEATS method, updated each February, with the new factors used to revise the previous five years of seasonally adjusted data.

For January 2026, revised seasonal factors and seasonally adjusted indexes for 2021 to 2025 were calculated and published. Seasonally adjusted indexes older than the last five years are considered final.

BLS is direct about the consequence. It advises against the use of seasonally adjusted data in escalation agreements precisely because those series are revised annually for five years.

There is a further detail that almost nothing outside the technical note mentions. Not every part of the index is seasonally adjusted, and the mix is reviewed each year against statistical criteria.

For 2026, 36 of the 81 components of the US city average all items index are not seasonally adjusted. The headline adjusted figure is therefore an aggregation in which a substantial minority of components carry no adjustment at all.

BLS also applies intervention analysis to series distorted by unusual events, estimating and removing those distortions before the seasonal factors are calculated. For the factors introduced for January 2026 it adjusted 57 series that way, including selected food and beverage items, motor fuels and vehicles.

How Precise the Monthly Change Actually Is

The index is a statistical estimate subject to sampling error, because it rests on a sample of retail prices rather than the complete universe of all prices. BLS says so plainly and publishes the size of that error.

The estimated standard error of the 1-month percent change is 0.04 per cent for the US all items CPI. BLS calculates and publishes standard errors for the 1-month, 2-month, 6-month and 12-month percent changes annually for the CPI-U.

The worked example is the useful part. For a 1-month change of 0.2 per cent in the all items CPI-U, BLS states 95 per cent confidence that the actual change based on all retail prices would fall between 0.12 and 0.28 per cent.

The market reaction to a small surprise is still real, because participants trade the difference against expectations. What the band changes is the inference: a narrow beat is information about positioning rather than about prices.

The transmission from inflation data to a currency runs through rate expectations, which is the subject of our guide to central bank policy stance rather than of the index itself. That transmission is not guaranteed: there is a regime in which a high inflation print stops being currency-positive, because the tightening it would normally imply may not be available.

Index Points Are Not Percentages, and Cities Are Not Comparable

Two misreadings appear often enough to be worth stating explicitly, and BLS addresses both.

The first is treating index point movements as percentages. Movements are usually expressed as percent changes rather than index points because point changes are affected by the level of the index relative to its base period, while percent changes are not.

An 18 point rise on an index at 225 is an 8 per cent change; the same 18 points on an index at 110 is 16.4 per cent. The point total alone carries no meaning without the level it moved from.

The second is comparing cities. BLS publishes indexes for 23 selected local areas, and states that area indexes do not measure differences in the level of prices among cities.

They only measure the average change in prices for each area since the base period. A city with a higher index number is not a more expensive city, and using the figures that way answers a question the data was never built to answer.

Where the release sits alongside other scheduled data is covered in our guide to the economic calendar, and the equivalent measures for another economy in our guide to other economic indicators.

Who This Page Is Not For

This page gives no view on where inflation is heading and no expectation for any forthcoming release.

It contains no method for trading the report, no levels, and no rule connecting a beat or a miss to a currency direction.

It is written for a reader who wants to know how the number is produced, how precise it is, and which version of it will still say the same thing in three years.

Frequently Asked Questions