TL;DR. The Federal Reserve targets 2 percent inflation using the Personal Consumption Expenditures (PCE) price index published by the Bureau of Economic Analysis, not the more familiar Consumer Price Index (CPI) from the Bureau of Labor Statistics. The two measure a different basket, use a different formula, and can tell different stories in the same month. In July 2026, headline CPI ran 3.4% year over year while headline PCE ran 3.7% — the gap that lands on the FOMC’s table is not the one dominating cable-news chyrons.
The two indexes at a glance
CPI and PCE both try to answer the same question: how much more do consumers pay this month than they did a year ago for a comparable bundle of goods and services? They arrive at different answers because they define the bundle differently and use different math.
| Attribute | CPI-U | PCE Price Index |
|---|---|---|
| Publisher | Bureau of Labor Statistics (BLS) | Bureau of Economic Analysis (BEA) |
| Formula | Modified Laspeyres (geometric mean within most categories) | Fisher-Ideal (chain-weighted) |
| Scope of spending | Out-of-pocket purchases by urban households (~93% of the U.S. population) | Spending by and on behalf of all households plus nonprofit institutions serving households |
| Weight source | Consumer Expenditure Survey (households) | Business surveys inside the National Income and Product Accounts (NIPA) |
| Weight refresh | Annually since 2023 (biennial before that) | Every month (chain-weighted with current-period expenditure data) |
| Release cadence | Monthly, typically mid-month | Monthly, typically late in the following month |
| Fed’s role | Cross-check | The 2% target the FOMC has used since 2000 |
The four real differences — and why they matter
1. Formula: Laspeyres vs. Fisher
CPI is a modified Laspeyres index. That means the BLS holds the basket roughly fixed and asks, “what does last year’s basket cost this year?” PCE uses a Fisher-Ideal formula that averages the base-period basket and the current-period basket. In plain English: if beef gets expensive and shoppers switch to chicken, the Fisher formula picks up the substitution almost immediately. Laspeyres does not until the next weight update. All else equal, that one difference makes CPI print a bit hotter than PCE in periods of active substitution.
2. Scope: out-of-pocket vs. everything spent on your behalf
This is the largest single reason PCE and CPI can look different. CPI counts what a household actually pays out of its own wallet. PCE counts spending “by and on behalf of” households — so employer-paid health insurance, Medicare, Medicaid, and payments by nonprofits all land in PCE. Because health care is a huge line item and most of it is paid by someone other than the patient, health services carry a much larger weight in PCE than in CPI. When medical inflation runs hot, PCE feels it more than CPI.
3. Weight sources: businesses vs. households
CPI weights come from the Consumer Expenditure Survey — households recording what they bought. PCE weights come from the supply side: business surveys, industry data, and administrative records feeding the NIPA accounts. Surveys of households tend to under-report certain categories (alcohol, tobacco, dining out). Business data tends to capture the full receipt. That is another quiet reason the weight vectors diverge.
4. Refresh cadence
PCE is chain-weighted, so its basket effectively updates every month. CPI weights are now updated annually (they were biennial before 2023 and roughly once per decade before the 1980s). During fast-moving inflation regimes, a slower-moving basket is a slower-adapting measure.
Why the Fed picked PCE
The FOMC did not always target PCE. In its February 17, 2000 Monetary Policy Report to the Congress, the Committee said it was changing its primary measure of inflation from CPI to the chain-type price index for personal consumption expenditures. The stated logic sits on three legs: the chain-weighted formula adapts to substitution automatically, the scope covers a wider slice of household consumption, and the underlying data can be revised as more information arrives, which the Fed views as a feature (better information) rather than a bug. The 2% goal itself was made explicit later, in the FOMC’s January 2012 statement on longer-run goals.
A worked example: substitution bias in one aisle
Imagine last year the “average” shopper bought 10 pounds of ground beef at $5 and 5 pounds of chicken at $4. Total: $70.
- This year beef jumps to $7 and chicken to $4.20. Shoppers respond: they buy 6 pounds of beef and 9 pounds of chicken.
- Laspeyres (CPI-style): price last year’s basket at this year’s prices. 10 × $7 + 5 × $4.20 = $91. Inflation reads (91 / 70) − 1 = +30.0%.
- Fisher (PCE-style): geometric mean of Laspeyres and Paasche (current basket at both years’ prices). Paasche = (6 × $7 + 9 × $4.20) / (6 × $5 + 9 × $4) = 79.80 / 66 = +20.9%. Fisher ≈ √(1.30 × 1.209) − 1 ≈ +25.4%.
Same reality, different numbers, because one measure assumes shoppers kept buying the expensive thing and the other lets them switch. Real-world differences per month are tiny slivers of this, but they compound.
Where they diverge right now
The most recent readings for July 2026 tell the story cleanly:
Notice what the chart makes obvious: headline PCE is higher than headline CPI this month (3.7% vs 3.4%) — the opposite of the typical post-2000 pattern in which PCE has usually run a touch below CPI. In 2026, medical services and shelter dynamics have driven the two indexes apart in ways that make the FOMC’s preferred gauge look stickier than the one on the front page. That is what the Yahoo Finance headline “Sticky PCE inflation leaves a divided central bank ahead of Fed’s Jackson Hole retreat” was describing.
Common misinterpretations
- “The Fed targets 2% CPI.” No — the target is 2% PCE, and it has been since the FOMC’s 2012 longer-run goals statement.
- “Core CPI and core PCE mean the same thing.” Both strip food and energy, but the underlying baskets, weights, and formulas still differ, so the two “cores” will not match.
- “CPI runs hotter than PCE, always.” Historically true on average, but the gap flips regularly — especially when medical services, financial services, or imputed rent move differently than out-of-pocket categories.
- “Higher CPI automatically means bigger rate hikes.” Only if PCE agrees. The FOMC watches CPI as a cross-check, not the target.
How to read a data day like a pro
When BLS drops CPI mid-month, note the headline and core year-over-year, but treat them as an early read on what PCE will show two weeks later. Some CPI line items feed directly into PCE (they are used to price the same underlying goods), so the PCE surprise on release day is often smaller than the CPI surprise. Others — notably medical services, portfolio management fees, and airline fares — are constructed differently. Watch those categories inside the CPI release if you want to guess where PCE will diverge.
For Fed watchers, the sequence that matters is: CPI release → PPI release → component-level nowcast → PCE release → Fed speeches. Anyone quoting only CPI in the same sentence as “the Fed’s 2 percent target” is skipping a step.
What to learn next
- How core inflation is constructed and why the Fed also watches trimmed-mean and median PCE.
- Owner’s equivalent rent (OER) — the biggest single line in CPI and one of the most controversial imputations in official statistics.
- The Fed dot plot and how PCE forecasts drive it.
- Yield-curve pricing of inflation via TIPS breakevens versus survey-based expectations.
Sources
- Federal Reserve — Why does the Federal Reserve aim for inflation of 2 percent over the longer run?
- Bureau of Economic Analysis — Personal Consumption Expenditures Price Index
- BEA — Personal Income and Outlays, July 2026 (released August 26, 2026)
- Personal Consumption Expenditures Price Index — overview citing FOMC’s Feb 17, 2000 Monetary Policy Report
- United States Consumer Price Index — BLS methodology summary
- U.S. CPI, July 2026 — BLS release aggregated by Trading Economics
Disclosure: This article is for informational purposes only and is not investment advice.