CPI vs PCE inflation: what does core leave out?

Which purchases are included in the price measure?

A price index combines many prices into a number that can be compared with a reference period. The percentage increase in that index is an inflation rate. CPI here means the US Consumer Price Index for All Urban Consumers, or CPI-U, published by the Bureau of Labor Statistics (BLS). PCE means the US Personal Consumption Expenditures price index from the Bureau of Economic Analysis (BEA). Other countries’ consumer price indexes have their own coverage and methods. A change in PCE spending in money terms is also different from a change in the PCE price index.

Both indexes have a broad measure called headline and a reduced measure called core that excludes specified food and energy categories. Core is not a third index alongside CPI and PCE. The weight of an item means the share assigned to it when combining prices: a bigger weight gives its movement more influence. Excluding food and energy does not make every remaining price stable; rent, medical care and other services can still move sharply.

On a narrow screen, scroll the tables sideways to see every column.

The food distinction is easy to miss. A restaurant meal falls within CPI's food-away-from-home category, which core CPI excludes. PCE places restaurant services under services, outside its excluded food category. Saying that every core index removes every food-related purchase would therefore be misleading.

Different coverage and weights can produce different inflation rates

PCE includes health spending paid for households by employers or government programs. CPI’s weights instead reflect a different scope of household expenditure. A medical-price change can therefore have a different influence in the two totals. But deciding the spending weight and observing the actual medical price are separate steps: CPI’s observed price is not always just the patient’s cash payment.

If a category takes a larger share of the index, the same price increase in that category has more effect on the total. Housing and health care need not have the same shares in CPI and PCE. Even when the underlying price information is similar, different weights can produce different overall increases.

People can respond to a more expensive item by buying another one. This is substitution. PCE uses spending patterns from neighboring periods to measure price change and links the changes across periods; the method is called a chain-type Fisher calculation. CPI also updates spending weights, and many detailed item groups reflect substitution in their calculations. CPI is therefore not a shopping basket frozen forever while PCE alone follows changes in what people buy.

Later information can also change past numbers. PCE can be revised as the national accounts underlying it are updated. Unadjusted CPI series are generally final on release, while seasonal factors and seasonally adjusted series can be revised. Seasonal adjustment estimates and removes recurring yearly patterns, such as holiday spending, to help reveal other changes. When an old article differs from today’s data table, check both the series name and which data version each used.

Core inflation is recalculated after the exclusions

Consider three invented spending groups. Hold their quantities and quality fixed and give each a starting weight, or share of total spending. Multiply a group’s weight by its price change to find its contribution to the overall rate. Contribution is measured in percentage points, the amounts added to or subtracted from a percentage rate. The table isolates this simple fixed-weight arithmetic; it does not reproduce either official index or its actual categories.

The headline change is 2.10%: 0.20 × 10 + 0.10 × (−20) + 0.70 × 3 = 2.10. Excluding the first two groups leaves 70% of the starting weight. Reset that remaining group to 100% of the reduced index: 2.10 ÷ 0.70 = 3.00. The toy core change is 3.00%.

The food and energy contributions cancel in this particular table, leaving the other group’s contribution to be rescaled across the reduced index. In general, subtracting the 10% food rise or adding back the 20% energy fall would ignore the different spending shares. Even removing their weighted contributions is only an intermediate step: the weights of everything kept must then be reset to make up the entire reduced index. Actual CPI and PCE calculations contain much more detail.

Food and energy still affect a household's budget. Core measures help examine price movement outside their specified categories; they do not replace the headline measure or a family's own expenses. Core can also be below headline when excluded categories rise quickly.

The same monthly rise can come with a lower annual rate

A percentage change needs a starting value: (new index ÷ earlier index − 1) × 100. To measure one month, start with the previous month; to measure a year, start with the same month a year earlier. The following fictional series uses index points, not inflation percentages. Its dates are for the arithmetic, not actual US readings or forecasts, and it has no seasonal pattern. Do not fill in months that are not shown.

The latest monthly rise is 0.20% in both cases. Yet the annual comparison falls from 3.21% to 2.39%. October compares with a year-earlier level of 101, while September compares with 100, so October starts from a higher base. This change in the past comparison value is called a base effect. Current prices keep rising even as the measured annual rise gets smaller; slower inflation does not mean prices have returned to their old level.

Real releases can use different seasonal-adjustment conventions for different comparisons. A seasonally adjusted monthly CPI series and an unadjusted annual comparison may use different inputs under the published method. Our fictional series contains no seasonal variation.

Annualizing a monthly rate

If a 0.20% monthly increase repeated for twelve months, the compounded rise would be 2.43%: (1.00212 − 1) × 100. Simply multiplying by twelve gives 2.40%, which leaves out compounding.

Compounding applies each month’s increase to the level already raised by the earlier months. Repeating one month in this hypothetical way is annualization: expressing its pace over a year. The observed annual change instead compares actual index levels twelve months apart. Neither the simple nor compounded annualized rate is that observed change, and a single month does not establish that its pace will continue.

What an inflation release can tell you

The release identifies the index, headline or core coverage, comparison period, seasonal adjustment and revision status. Its category breakdown then explains what drove the change. A falling energy price and slower service inflation can produce similar totals with different explanations.

One inflation figure cannot promise a rate cut or an immediate change to a mortgage payment. Inflation is part of a broader policy assessment, and household borrowing costs do not all reset at once. When comparing countries, start with each country's index definition as well as its policy-rate framework.

A country’s production value can rise because it produces more or because prices rise. GDP adds up production value; nominal GDP includes price changes, while real GDP seeks to separate changes in the quantity produced. Our GDP vs GNI guide follows that distinction, income received across borders and conversion into per-person amounts in different currencies.

Sources and assumptions

• BLS: CPI coverage, methods, seasonal adjustment and revisions
• BLS: CPI item aggregation
• BEA: differences between CPI and PCE
• BEA: core PCE exclusions
• BLS: medical care measurement
• BEA: PCE revisions and changing consumer spending

The links identify the specific BLS and BEA methods discussed above. Wording and all example tables were prepared for this guide; no official series, chart, photograph or logo is reproduced.

Every number, weight and dated index level in the examples is fictional, including the dates within the series. They are not actual CPI or PCE observations, forecasts or official expenditure shares. The three-group model uses fixed starting weights solely to explain contributions and renormalization; it does not reproduce CPI or PCE methods or category boundaries. Period comparisons use full precision before rounding.

This guide explains statistical concepts and does not recommend investments or predict interest-rate decisions. See our editorial policy.