On a Friday morning when both price reports are out, an illustrative example, a reader sees one percent change under a Bureau of Labor Statistics headline and a different percent change under a Bureau of Economic Analysis headline and assumes one office has corrected the other. Neither office is grading the other. The Consumer Price Index and the Personal Consumption Expenditures price index are built for overlapping but not identical questions. They can move together for years and still print different numbers in a given month for reasons the agencies have described in public.

Two products, two agencies

The CPI comes from BLS. The broad series in news briefs is the all-items CPI for All Urban Consumers, the CPI-U. BLS defines it as a measure of the average change over time in prices paid by urban consumers for a market basket of goods and services. The PCE price index comes from BEA, inside the national accounts. It prices the goods and services consumed by households, and by nonprofit institutions serving households, including some spending that a household does not pay out of pocket.

The Federal Reserve has said which of the two it uses as its longer-run yardstick. In the 2025 Statement on Longer-Run Goals and Monetary Policy Strategy, the Federal Open Market Committee reaffirmed that inflation at 2 percent, measured by the annual change in the price index for personal consumption expenditures, is most consistent over the longer run with its employment and price-stability mandates. The Board's own inflation explainer adds that the PCE index adapts more quickly to changes in spending patterns. A headline about "the Fed's inflation target" is a PCE sentence. A headline about Social Security's cost-of-living adjustment is a CPI-W sentence. Treating every percent sign as the same object is how the disagreement starts.

Four effects the agencies already named

BLS economist Noah Johnson, in a 2017 research paper, grouped CPI–PCE differences into formula, weight, scope, and a residual category he called "other." An earlier BLS Beyond the Numbers article on the same gap walks through the first three in operational detail. You do not need to pick a winner to use the list. You need the list so a one-month gap has a place to go.

Formula. At the upper level, CPI-U and CPI-W aggregate basic indexes with a modified Laspeyres formula. The CPI chapter of the BLS Handbook of Methods says the quantity weights stay fixed for a year and are replaced each January. In that aggregation, substitution between items is taken to be zero. PCE uses a Fisher ideal index, which BEA can compute because it rebuilds weights as new expenditure data arrive. A Fisher index reflects substitution among detailed items when relative prices change. BLS has noted that a superlative index is hard to run in real time, because it wants expenditure data for the period that is still unfolding. The chained CPI-U is BLS's own superlative-style series, and it is revised; the headline CPI-U is not.

Since January 1999, BLS has also used a geometric mean for most basic indexes inside item categories, which allows a modest amount of substitution within a category such as types of ground beef in one city. That lower-level choice is not the same thing as the upper-level Laspeyres aggregation, and it is not the PCE Fisher index. A sentence that says "the CPI assumes you switch from steak to hamburger" is the misconception BLS has answered directly: steak and ground beef sit in different item categories, so the CPI-U does not build that swap in. The formula gap with PCE is real, and it is narrower than the slogan.

Weights. CPI weights come chiefly from the Consumer Expenditure Survey, a household survey. PCE weights come from business surveys that feed the national accounts, including retail trade surveys and service surveys. Households and stores do not describe the same spending with the same error, the same timing, or the same classification. Even a category with the same English name can carry a different share in the two baskets. Weight differences move the indexes when prices in heavily weighted categories diverge.

Scope. CPI-U measures out-of-pocket spending by urban households. PCE measures consumption by all households, plus nonprofits serving them, including rural households and spending made on a household's behalf. Employer-paid health insurance and government programs such as Medicare and Medicaid are the example both agencies reach for: those payments are largely outside the CPI medical basket and inside PCE. A year when medical prices and medical coverage shift will open a gap that has nothing to do with anyone "cooking the books." It is a different boundary around the word "consume."

Other. Classification choices, how a component index is built, and timing can leave a remainder after formula, weight, and scope are accounted for. Johnson's point, echoing earlier BLS work by Fixler and Jaditz, is that the gap is an accumulation of small design differences. There is no single switch that explains every month.

What "both right" can honestly mean

Right means "faithful to the question the index was built to answer." CPI-U answers: how did out-of-pocket prices change for the average urban household, on a formula that contracts and tax parameters can use because it is final when published? PCE answers: how did the prices of household consumption in the national accounts change, with a superlative formula and a broader scope, including third-party payments? The Fed's 2 percent goal attaches to the second question. Escalation clauses, TIPS, and the CPI-W benefit formula attach to versions of the first.

A complete cost-of-living index is a third question, and BLS says the CPI is not that index. A full cost-of-living measure would track the spending required to hold a standard of living constant, including factors the CPI does not quantify. Calling either published percent "the cost of living" overclaims both of them.

Core indexes add a fourth, narrower question. All-items CPI-U includes food and energy. All-items less food and energy leaves them out because those prices are volatile and analysts want a smoother read on the rest of the basket. BLS is explicit that food and energy are still inside the headline CPI, and that the legislated uses it lists — Social Security, federal retirement, income-tax parameters, TIPS — use all-items indexes, not the core. A news line about "core inflation" has dropped part of the basket on purpose. It has not replaced the headline.

How to read a month when the numbers split

Write the two figures down with their names before you explain the split. Then walk the four effects in order, and stop when you have said enough to orient the month.

  1. Confirm you are comparing the same span. A one-month change, a 12-month change, and a quarterly annualized change are three sentences.
  2. Confirm you are comparing the same concept of "core" or "all items." PCE excluding food and energy against headline CPI-U is a double difference.
  3. Ask whether medical care or another third-party-heavy category was a large mover. That is a scope candidate.
  4. Ask whether households were shifting spending across categories. PCE will pick that up faster than a CPI-U still using its current annual weights. That is a formula and weight candidate.
  5. If a contract or a benefit names one series, the other series is context, not a correction. The clause already chose.

The calculator article is the practical companion: most public calculators default to CPI-U all items, not to PCE, and they will not warn you when a headline you saw that morning was the other index. Wage comparisons have a further choice, because BLS deflates some earnings series with CPI-U and others with CPI-W. That choice is part of wage growth that outruns prices.

Shelter is a special case inside the scope discussion. CPI shelter is mostly rent and owners' equivalent rent, not the sale price of houses. PCE housing is its own national-accounts concept. Putting a house-price index next to either one and expecting them to match is the mix-up traced in house prices, rent, and the same number trap.

A table you can keep next to a headline

CPI-U (BLS) PCE price index (BEA)
Question Out-of-pocket prices for urban consumers Prices of household consumption in the accounts
Upper-level formula Modified Laspeyres, weights fixed within the year Fisher ideal, chained as data arrive
Weight source Household expenditure survey Business and administrative source data
Medical scope Out-of-pocket spending Also spending by employers and government programs on behalf of households
Typical use Escalation, TIPS, tax parameters (CPI-W for Social Security) Federal Reserve's longer-run inflation gauge

The cells are a reading guide, not a ranking. Sources for the rows are the BLS handbook and the BLS comparison of the two indexes, Johnson's 2017 paper, and the Federal Reserve's strategy statement. Definitions of the plain-language terms live in the glossary.

What this page will not do

Nothing here tells you which index will be higher next quarter, or which one to use in a private contract that does not already name one. Design details are updated by BLS and BEA; the papers cited above describe the structure, and the latest handbook chapter is the place to confirm a formula before you rely on it. This is educational context, not investment advice and not a prediction.

When the next pair of headlines disagrees, which question were you actually trying to answer, the urban out-of-pocket basket or the national-accounts consumption basket?