Measuring Inflation: CPI and PCE

How the two main US price indexes are built: the CPI's sample of prices and its spending weights, how housing is measured and why it lags, quality changes and substitution, the chained CPI, and the four reasons the PCE price index has usually risen more slowly than the CPI.

Lesson 6 of 13 · 8 min read · Advanced · Free · Last reviewed October 2026

The short version

The Consumer Price Index tracks about 80,000 prices a month, weighted by what urban households spent two years earlier. Shelter is over a third of it, measured mostly through rents that change slowly, so shelter inflation lags the market: in the last inflation wave it peaked nine months after the headline CPI. A fixed basket misses people switching to cheaper goods, and the chained CPI, which allows for that, rose 75% from 1999 to 2024 against 87.5% for the regular CPI. The PCE price index, the one the Federal Reserve targets, uses a different formula, different weights, and a wider scope, and has risen about 0.4 percentage point a year less than the CPI.

By the end of this lesson, you'll be able to

  • Describe how the CPI is built: the sample of prices, the spending weights, and the treatment of housing
  • Explain quality adjustment and substitution, and why the chained CPI rises more slowly
  • Explain the four sources of difference between the CPI and the PCE price index

Why this matters

A price index turns millions of individual prices into one number, and that number does real work. The principal of a TIPS bond is adjusted using the CPI,1 the Federal Reserve's 2% goal is defined with the PCE price index,2 and every "real" return in this course subtracts one of them. The choices made in building an index, about what to include, how to weight it, and how to handle better products, change the answer by enough to matter over a few decades.

How the CPI is built

Each month BLS records the prices of about 80,000 items, classified into more than 200 categories.3 The main index, the CPI-U, covers urban consumers, who make up over 90% of the US population. A narrower version, the CPI-W, covers households of wage earners and clerical workers, about 30% of the population.3

Prices are combined using spending weights from the Consumer Expenditure Surveys, which ask households what they buy. Since 2023, BLS has updated the weights every year, using spending from two years earlier.4 An item's weight, called its , is its share of total spending in the index.

What the CPI-U is made of, December 2025
ComponentShare of the indexLargest parts
Shelter35.6%Owners' equivalent rent 26.2%, rent 7.8%
Other services25.1%Medical care services 6.9%, transportation services 6.3%
Goods other than food and energy19.2%New vehicles 3.8%, used cars and trucks 2.8%, apparel 2.4%
Food13.7%At home 8.3%, away from home 5.4%
Energy6.4%

Bureau of Labor Statistics figures.5 Services make up most of the index, and shelter alone is more than a third. Food and energy together are about a fifth.

From weights to the headline

Because the index is a weighted average, each component's contribution to inflation is roughly its weight times its own price change.

Contribution ≈ Weight × Price change of the component

Worked example: building a year's inflation from its parts (hypothetical price changes)
ComponentWeightPrice changeContribution
Shelter35.6%+3.0%+1.07 points
Other services25.1%+3.5%+0.88 points
Goods other than food and energy19.2%+1.0%+0.19 points
Food13.7%+2.5%+0.34 points
Energy6.4%−10.0%−0.64 points
All items100%+1.84%

A 10% fall in energy prices takes only about 0.6 point off the total, while a modest 3% rise in shelter adds more than a point. The weights shift a little as relative prices change, so this is an approximation.

Housing: the biggest and slowest part

The CPI doesn't price owned homes. BLS treats a house as a capital, or investment, good, separate from the shelter it provides, and measures the cost of that shelter as : what the home would rent for.6 House prices themselves aren't in the index.

Both rent and owners' equivalent rent come from a survey of rental units. Each unit in the sample is priced every six months, and the monthly change is worked out from the six-month change. Many rents are locked in for the length of a lease.6 The result is that shelter inflation in the CPI moves slowly and late.

CPI inflation over 12 months, all items and shelter
MonthAll itemsShelter
June 20229.1%5.6%
March 20235.0%8.2%
June 20243.0%5.1%
August 20263.4%3.0%

Bureau of Labor Statistics figures.7 Headline inflation peaked in June 2022. Shelter kept climbing for nine more months and peaked in March 2023, by which time headline inflation had fallen to 5%. Because each unit is repriced only every six months and many rents are fixed for a lease, a change in rents for new leases reaches the index gradually.

Better products and changing habits

Quality change

When an item in the sample is replaced by a newer model, part of any price difference may be paying for a better product. BLS adjusts for this, in some cases with statistical analysis of how much each feature is worth.3

Worked example: a laptop replaced by a newer model (hypothetical)
Old model's price$1,000
New model's price$1,050
Estimated value of the improvements$40
Price change counted$10, or 1%

Without the adjustment, the index would record a 5% rise for something that is partly a better machine. The size of the adjustment rests on an estimate, which is one reason measured inflation is never exact.

Substitution and the chained CPI

The regular CPI prices a fixed basket. When one item gets expensive, people often buy less of it and more of something similar, and a fixed basket misses that switch, so it can overstate the rise in the cost of keeping the same standard of living. BLS publishes a chained CPI, the C-CPI-U, that allows for substitution across categories. Its final figures need spending data that aren't available until 10 to 12 months after the month they describe.3

Over long periods the gap adds up. From December 1999 to December 2024, the CPI-U rose 87.5%, about 2.55% a year, while the chained CPI rose 75.2%, about 2.27% a year.8

CPI and PCE: four differences

The PCE price index, from BEA, measures prices across personal consumption expenditures in the national accounts. A BEA study comparing the two found four sources of difference:9

  • Formula. The CPI uses a modified Laspeyres formula, which works from a fixed basket. The PCE price index uses a Fisher-Ideal formula, which uses spending patterns from both periods it compares, so it allows for substitution.
  • Weights. CPI weights come mainly from household surveys, PCE weights mainly from business surveys. Housing is the biggest difference: at the time of the study, rent of shelter was about 32% of the CPI and about 15% of the PCE index.
  • Scope. The CPI covers what urban households pay out of pocket. The PCE index also covers spending made on households' behalf, such as medical care paid for by employer health insurance or by Medicare and Medicaid.
  • Other. Smaller differences in seasonal adjustment and in the prices used.

From 2002 to mid-2007, the period studied, the CPI rose about 0.4 percentage point a year faster than the PCE index, and the formula difference explained almost half of the gap.9 The gap has held over longer periods. From December 1999 to December 2024, the CPI rose about 2.56% a year and the PCE index about 2.19%. For the core measures, without food and energy, the figures were 2.40% and 2.09%.101112 It isn't a fixed rule, though: over the 12 months to August 2026, core PCE rose faster than core CPI, as Reading Economic Data Releases showed.

Which index to use

Each index answers a slightly different question, and the right one depends on the use.

Where each measure is used
UseIndex
The Federal Reserve's 2% inflation goalPCE price index2
Adjusting the principal of TIPSCPI-U, not seasonally adjusted13
Breakeven inflation in the bond marketExpected CPI inflation, since it's built from TIPS yields
Your own costsNeither exactly: your spending isn't the average basket

When you compare an investment's return with inflation, the choice of index can move the answer by a few tenths of a point a year, and over 25 years that compounds into a real difference: 87.5% against 75.2% for the two versions of the CPI alone.

Key takeaways

  • The CPI tracks about 80,000 prices a month, weighted by household spending from two years earlier, updated yearly.
  • A component's contribution to inflation is roughly its weight times its price change. Shelter is over a third of the CPI.
  • Housing enters as rent and owners' equivalent rent, which move slowly. Shelter inflation peaked nine months after the headline in 2023.
  • Quality adjustment removes the part of a price rise that pays for a better product. The chained CPI allows for substitution and rises more slowly.
  • The PCE price index differs from the CPI in formula, weights, and scope, and has usually risen about 0.3 to 0.4 point a year less.
  • The Fed targets PCE inflation. TIPS are adjusted with the CPI-U.

Check your understanding

Energy is about 6.4% of the CPI. If energy prices rise 15% over a year and nothing else changes, roughly how much does that add to 12-month CPI inflation?

Why has the chained CPI usually risen more slowly than the regular CPI?

What's next

Measuring inflation is one question. Explaining it is another. What Drives Inflation looks at the causes: demand, supply shocks, money, and expectations.

Go deeper

Notes and sources

  1. TreasuryDirect, U.S. Department of the Treasury. Treasury Inflation-Protected Securities (TIPS). Checked 2026-10-14. Back to text
  2. Board of Governors of the Federal Reserve System. Why does the Federal Reserve aim for inflation of 2 percent over the longer run?. Checked 2026-10-14. Back to text
  3. U.S. Bureau of Labor Statistics. Consumer Price Index: Frequently Asked Questions. Checked 2026-10-14. Back to text
  4. U.S. Bureau of Labor Statistics. Relative Importance and Weight Information for the Consumer Price Indexes. Checked 2026-10-14. Back to text
  5. U.S. Bureau of Labor Statistics. Relative importance of components in the Consumer Price Indexes: U.S. city average, December 2025. Checked 2026-10-14. Back to text
  6. U.S. Bureau of Labor Statistics. Measuring Price Change in the CPI: Rent and Rental Equivalence. Checked 2026-10-14. Back to text
  7. Federal Reserve Bank of St. Louis (FRED), from the U.S. Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers: All Items (CPIAUCNS) and Shelter (CUSR0000SAH1) in U.S. City Average, monthly. Checked 2026-10-14. Back to text
  8. Federal Reserve Bank of St. Louis (FRED), from the U.S. Bureau of Labor Statistics. Chained Consumer Price Index for All Urban Consumers: All Items (SUUR0000SA0) and Consumer Price Index for All Urban Consumers: All Items (CPIAUCNS), U.S. City Average, monthly. Checked 2026-10-14. Back to text
  9. U.S. Bureau of Economic Analysis. Clinton P. McCully, Brian C. Moyer, and Kenneth J. Stewart, Comparing the Consumer Price Index and the Personal Consumption Expenditures Price Index, Survey of Current Business (November 2007). Checked 2026-10-14. Back to text
  10. Federal Reserve Bank of St. Louis (FRED), from the U.S. Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers: All Items in U.S. City Average (CPIAUCSL), monthly. Checked 2026-10-14. Back to text
  11. Federal Reserve Bank of St. Louis (FRED), from the U.S. Bureau of Economic Analysis. Personal Consumption Expenditures: Chain-type Price Index (PCEPI) and excluding food and energy (PCEPILFE), monthly. Checked 2026-10-14. Back to text
  12. Federal Reserve Bank of St. Louis (FRED), from the U.S. Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers: All Items Less Food and Energy in U.S. City Average (CPILFESL), monthly. Checked 2026-10-14. Back to text
  13. Code of Federal Regulations, via Cornell Law School Legal Information Institute. Appendix B to 31 CFR Part 356: Formulas and Tables. Checked 2026-10-12. Back to text

Saved to this browser in the prototype. On the live site, progress saves to your account.

This lesson is general education, not personalized investment, tax, or legal advice. Examples use fictional people and companies, and hypothetical numbers.