The Bureau of Labor Statistics (BLS) calculates the Consumer Price Index every month. It prices a sample of goods, services and rents, turns those prices into 7,776 small item-by-area indexes (243 items in 32 index areas), and combines them with weights from household spending surveys. The inflation rate you see in a headline is then one division: the index for this month over the index for the comparison month, minus one, times 100.

Last checked November 12, 2026, against the BLS handbook, CPI notices and the October 2026 CPI release. Our US economic calendar tracks the release schedule; this guide explains what the release measures and how to redo its arithmetic.

How to do the CPI calculation yourself

The percent change between two months is:

(index in the later month ÷ index in the earlier month − 1) × 100

Use index levels from the same series, such as CPI-U all items, not seasonally adjusted, 1982-84=100. These are the published index levels for that series:

Month CPI-U index level
September 2025 324.800
October 2025 not published
November 2025 324.122
September 2026 336.940
October 2026 337.511

The 12-month change to September 2026 is (336.940 ÷ 324.800 − 1) × 100 = 3.7%.

October 2026 is the awkward month. BLS did not collect survey data for October 2025 during the lapse in appropriations, so for most series there is no October 2025 index, and the October 2026 release says BLS cannot publish 12-month changes where that base month is missing. Dividing by November 2025 instead gives 4.1%, but that is an 11-month change, not an annual rate.

BLS’s factsheet on missing data points describes a stopgap that contract parties can agree to use: take the geometric mean of the months on either side. Here, √(324.800 × 324.122) = 324.461, and (337.511 ÷ 324.461 − 1) × 100 = 4.0%. BLS is explicit that a figure built this way is unofficial and is not a CPI series.

The basket starts with spending, not a shopping list

The CPI basket represents goods and services bought for consumption. Food, housing, transport and medical care belong in it. Purchases of stocks and bonds do not: those are investments rather than consumption. Sales taxes associated with a purchase can enter its price, while income tax is outside the basket.

BLS uses its Consumer Expenditure Surveys to establish the importance of different spending categories. A household that spends more on rent than on haircuts does not give those purchases equal influence. The national index applies the same principle across the spending patterns of the population it represents.

This also explains why a person’s experience can differ from the headline. A renter with a long commute and a homeowner who walks to work have different exposures to shelter and fuel. The index describes an average population, not either household’s budget.

Price collection follows specific products

A price quote records a particular product or service at a selected outlet. Comparing a small pack last month with a larger pack this month without accounting for the difference would mix a price change with a quantity change.

According to the BLS Handbook of Methods, the survey collects about 100,000 prices a month for commodities and services and about 8,000 rental housing quotes. Not every price is checked every month. Food at home, energy and some other items are priced monthly, as is everything in New York, Los Angeles and Chicago; elsewhere most other items are priced every other month. The rent sample is split into six panels, and each unit is priced twice a year.

The calculation framework divides spending into 211 item strata and uses 243 basic items across 32 index areas, producing the 7,776 item-area combinations. A few strata, including airline fares and used vehicles, use secondary data sources, and price movements for 24 strata with small weights are imputed from related strata.

Collection is thinner than it was before 2025

In April 2025 BLS suspended CPI collection entirely in Lincoln, Nebraska, and Provo, Utah, and in June it did the same in Buffalo, New York. BLS reported that roughly 15% of the sample in the other 72 areas was also suspended, on average. Its simulation found the effect on the all-items 12-month change to be under 1/100th of a percentage point on average, while warning that local and item indexes may be more volatile.

When a price is missing, BLS first imputes from the same item in the same area, then from the same item in a wider region, and only then carries the last price forward, as its imputation page explains. In an October 22, 2026 update, BLS said it resumed hiring CPI field staff in November 2025, expects collection to resume for suspended parts of the sample, and has published monthly response and collection rates since June 2026.

Basic indexes and expenditure weights do different jobs

Most commodity and service categories use a geometric mean formula at the basic-index stage, an approach BLS has used for most basic indexes since January 1999. That formula combines price ratios for the sampled items. A limited set of categories, including electricity, prescription drugs and hospital services, uses a modified Laspeyres formula instead.

At the next stage, BLS aggregates basic indexes using expenditure weights. For CPI-U and CPI-W, those aggregation weights come from the Consumer Expenditure Surveys and are updated annually. The weights decide how much each component contributes to the aggregate.

Consider an illustrative, simplified budget: housing has a 50% weight, food 30% and other purchases 20%. If their prices rise 4%, 2% and 1%, respectively, a simple fixed-weight calculation gives 2.8%: 0.50 × 4 + 0.30 × 2 + 0.20 × 1. These are invented teaching inputs, not actual CPI weights or the complete BLS formula. The example shows why a large component can dominate the result even when a smaller component rises faster.

The 2025 shutdown reaches the weights too. BLS says the Census Bureau did not collect Consumer Expenditure data in October and November 2025, which affects the 2027 CPI-U and CPI-W weights, and that it chose a survey weight adjustment approach to keep those releases on schedule.

Replacements require a quality comparison

Products disappear or acquire new specifications. BLS must decide whether a replacement differs in ways that affect the measured purchase. Otherwise, a more expensive product with added features would automatically count as pure inflation.

The calculation handbook describes three treatments: direct comparison when the items are essentially the same, direct quality adjustment that estimates the value of the difference, and imputation when neither works. Quality adjustment aims to isolate price change for comparable consumption; it does not say that every buyer values an added feature equally.

Shrinkflation raises the related quantity question. If the package contains less while its ticket price stays unchanged, the price per equivalent quantity has increased. Looking only at the shelf label misses that change.

Read the series and the comparison period together

CPI-U covers urban consumers, more than 90% of the US population; CPI-W covers urban wage earners and clerical workers, about 30%. The chained CPI-U uses a different upper-level aggregation approach that accounts for substitution across categories. Its final expenditure data arrive 10 to 12 months after the reference month, so its initial estimates are revised.

The percentage change also needs a denominator. If an index rises from 300 to 309, the increase is 3%, calculated as (309 ÷ 300 − 1) × 100. A nine-point index increase is not nine percent.

Monthly seasonally adjusted changes and unadjusted year-over-year changes answer different questions. Seasonal adjustment removes recurring patterns to make adjacent months easier to compare; a twelve-month comparison looks across a much longer interval. Keep the series, adjustment basis and comparison period consistent when comparing releases.