Understanding the Consumer Price Index

The CPI measures the average change over time in the prices urban consumers pay for a basket of goods and services. The BLS prices the same kinds of items month after month, so that movements in the index reflect changes in prices rather than changes in what people buy. The basket's weights come from the Consumer Expenditure Survey and, since 2023, are updated every year.

The headline index is the CPI-U, which BLS says covers over 90% of the US population. The CPI-W covers households of urban wage earners and clerical workers, about 30% of the population. Social Security raises benefits each year by the change in the CPI-W's July–September average from the year before.

BLS data collectors record the prices of about 80,000 items each month at roughly 23,000 stores, service providers, and other outlets in 75 urban areas. A separate housing survey collects about 8,000 rent quotes a month. The individual price changes are averaged using weights that reflect how much urban consumers spend on each category.

Most CPI series use a 1982–84 reference base, set equal to 100. The CPI-U averaged 332.2 over January–September 2026, which means the same basket costs about 3.3 times what it did in 1982–84. The inflation rate is the percentage change in the index over a year. The monthly headline compares one month with the same month a year earlier. This site, like many long-run comparisons, uses annual averages instead: the CPI-U averaged 304.7 in 2023 and 313.7 in 2024, an increase of 2.9%.

BLS sorts the basket into eight major groups. The shares below are the CPI-U relative importance figures for December 2025, the latest annual table, which are based on 2024 spending:

Housing (44.5%): Shelter (35.6%), which is rent of primary residence (7.8%), owners' equivalent rent (26.2%), and hotels and other lodging away from home, plus household energy, water, furnishings, and operations.

Transportation (16.3%): New and used vehicles, motor fuel, maintenance, vehicle insurance, and public transportation including airfares.

Food and beverages (14.5%): Groceries, restaurant meals, and alcoholic drinks.

Medical care (8.4%): Prescription drugs, doctor and hospital services, and health insurance.

The other four are education and communication (5.8%), recreation (5.1%), other goods and services (2.9%), and apparel (2.4%). Shelter's share has grown: it was 33.3% in December 2020.

Owners' equivalent rent (OER) is the largest single item in the CPI, at 26.2% of the index. The BLS treats an owned home as an investment asset, so house prices and mortgage interest are not in the CPI. It prices the shelter service a home provides instead: homeowners' estimates of what their home would rent for set the weight, and the month-to-month change comes from rents on comparable rental units in the housing survey.

That design makes shelter slow to respond. The housing sample is split into six panels, and each unit is repriced only twice a year. Most tenants also renew leases annually, so CPI rents track what all renters pay, not the asking rent on new leases. When market rents jumped in 2021 and 2022, CPI shelter inflation rose gradually and kept rising after new-lease rents had leveled off; it then stayed elevated through 2024. The same design means the rise in mortgage rates in 2022 and 2023, which raised the monthly cost of buying a home, never entered the index directly.

Headline CPI includes every item. Core CPI leaves out food and energy, which together make up about a fifth of the index (energy alone is 6.4%). A harvest failure or an oil supply disruption can move those prices sharply within a few months and then reverse, so core inflation is a steadier read on the underlying trend.

The Federal Reserve defines its 2% inflation goal using the price index for personal consumption expenditures (PCE), produced by the Bureau of Economic Analysis. PCE covers a wider range of spending, including medical care paid for by employers and government programs on a household's behalf, so shelter carries far less weight in it. CPI inflation has usually run somewhat above PCE inflation. A 2013 analysis by the San Francisco Fed found that most of the gap came from the different weights, with shelter at 32% of the CPI against 15% of PCE, and only a small part from the formulas.

The standard CPI-U already allows for some substitution within narrow categories: since 1999 BLS has used a geometric mean formula for most of its basic indexes, so if one variety of apple gets expensive and shoppers buy another, the index reflects some of that. The Chained CPI (C-CPI-U), published since 2002, also captures substitution across categories, such as buying more chicken when beef prices rise. From 2001 to 2023 it rose about 0.2 percentage points a year less than the CPI-U, measured December to December. Its figures are revised for 10 to 12 months after first release. The Tax Cuts and Jobs Act of 2017 switched federal income tax bracket indexing to the chained CPI starting with the 2018 tax year.

The Producer Price Index (PPI), also published by the Bureau of Labor Statistics, measures the prices domestic producers receive for their output, including goods and services sold to other businesses. Higher input costs can reach consumers later, so analysts read the PPI as an early signal for the CPI. The link is loose: producers may absorb costs in their margins, and services such as shelter and medical care, which make up much of the CPI, barely appear in producer prices.

The CPI is an average across urban households. A renter in a city where rents are climbing fast faces a different inflation rate from a homeowner with a paid-off mortgage, whose cash housing costs are property tax, insurance, and repairs, even though owners' equivalent rent is a quarter of the index. BLS publishes separate indexes for regions and large metro areas, which narrow the gap but do not close it.

BLS also adjusts for changes in quality. When a phone model is replaced by a faster one at the same price, a statistical (hedonic) model estimates how much of that price pays for the better features, and the quality-adjusted price falls even though the sticker price did not. BLS uses such models for phones, apparel, major appliances, and rent, among other items. The method is sound in principle, but it means the index can show lower inflation than a shopper who simply compares receipts.

Finally, the CPI excludes asset prices. Rising home values, stock prices, and collectibles do not enter the index, so a first-time buyer facing higher house prices sees that cost reflected only indirectly, through rents.