CPI vs PCE
Last updated:
Direct answer
CPI (Consumer Price Index, BLS) and PCE (Personal Consumption Expenditures price index, BEA) both measure US consumer-price inflation, but they aren't the same number. PCE typically runs 0.3 percentage points BELOW CPI. The Federal Reserve targets 2% PCE, not 2% CPI — a fact frequently misquoted. The main differences are weighting (housing and health care), substitution treatment, and source data (household survey vs. business survey).
Why this question matters
Most public discussion of "inflation" cites CPI because BLS releases it first and headline media run with it. But the Federal Reserve's dual-mandate inflation target — locked at 2% since 2012 and reaffirmed in the August 2020 statement on longer-run goals — is denominated in PCE, not CPI. If you're modeling Fed reaction functions, pricing real-rate exposure, or backtesting macro signals, the choice between CPI and PCE materially changes the answer. The wedge isn't large in any single month, but it compounds: 0.3 percentage points over a decade is a non-trivial drift between the same nominal series rendered two ways. The two indices also disagree on what consumers actually spend money on, which matters more than the headline gap for sector-specific work (housing, health care, durable goods).
CPI vs PCE — side-by-side
| Dimension | CPI | PCE | Note |
|---|---|---|---|
| Publishing agency | Bureau of Labor Statistics (BLS) | Bureau of Economic Analysis (BEA) | Both US Department of Commerce / Labor; methodologies diverged historically |
| What it measures | Out-of-pocket household consumer prices | Total consumer expenditures including employer-paid + government-paid | PCE has broader scope by design |
| Source data | Consumer Expenditure Survey (households) | Business + administrative records (NIPA) | Household self-report vs business reporting + Medicare/Medicaid data |
| Formula | Modified Laspeyres (fixed basket, updated biennially) | Fisher Ideal (chain-weighted, updates continuously) | PCE captures substitution effects in near-real-time; CPI lags |
| Owner's-equivalent rent weight | ~24% | ~12% | CPI gives shelter twice the weight; major source of recent CPI-vs-PCE divergence |
| Health-care weight | ~7% | ~17% | PCE includes employer-paid premiums + Medicare/Medicaid; CPI is out-of-pocket only |
| Typical headline wedge | Reference | ~0.3pp below CPI on average (2010-2024) | Not constant; the wedge widens when housing inflates faster than health care |
| Release cadence | Monthly, ~10-15th of month for prior month | Monthly, ~last Friday for prior month | PCE incorporates the CPI data BLS releases two weeks earlier |
| Historical depth | Monthly back to 1913 | Monthly back to 1959 | CPI is the only US inflation series with century-plus depth |
| Federal Reserve target | Not the official target | 2% (since 2012, reaffirmed 2020) | Common misquote: media often say "Fed targets 2% inflation" implying CPI — it's PCE |
| Core variant (ex food + energy) | Core CPI | Core PCE | Fed Chair speeches more often cite Core PCE; markets watch both |
| Best for | Headline cost-of-living, COLA adjustments, TIPS pricing | Fed policy reaction modeling, macro nowcasts, GDP deflator linkage | TIPS reference CPI by indenture; PCE is the policy-relevant series |
Numbers worth remembering
- PCE runs roughly 0.3 percentage points below CPI on average over 2010-2024 (BLS / BEA series).
- Owner's-equivalent rent has ~2× the weight in CPI (~24%) versus PCE (~12%) — the single biggest source of recent divergence.
- Health-care services weight is ~7% in CPI versus ~17% in PCE, because PCE includes employer-paid premiums + Medicare/Medicaid.
- The Fed's 2% inflation target has been denominated in PCE (not CPI) since the January 2012 FOMC statement on longer-run goals, reaffirmed August 2020.
- CPI history goes back to monthly 1913 observations; PCE monthly history starts in 1959.
- CPI releases ~10-15th of each month for the prior month; PCE releases ~last Friday of each month — PCE incorporates the CPI release that preceded it by ~2 weeks.
How to pull this data
When the Tidore macro vertical ships its `/v1/macro/<indicator>` surface, the same comparison is one HTTP call for each series. Today the canonical primary sources are BLS and BEA directly.
# Tidore (in development, M3+ release) — pull both series from one API key
curl -H "Authorization: Bearer $TIDORE_API_KEY" \
"https://api.tidore.co/v1/macro/cpi?from=2010-01-01&to=2024-12-01"
curl -H "Authorization: Bearer $TIDORE_API_KEY" \
"https://api.tidore.co/v1/macro/pce?from=2010-01-01&to=2024-12-01"
# Today — pull from primary sources directly (no API key required)
# BLS CPI: https://www.bls.gov/cpi/data.htm
# BEA PCE: https://www.bea.gov/data/personal-consumption-expenditures-price-indexAuthoritative sources
- BLS — Consumer Price Index Overview
- BEA — Personal Consumption Expenditures Price Index
- FOMC — 2012 Statement on Longer-Run Goals and Monetary Policy Strategy (Fed PCE 2% target)
- Cleveland Fed — Difference Between CPI and PCE Inflation (research summary)
- BLS — Why does the BLS provide both the CPI and the PPI?
Frequently asked questions
Does the Federal Reserve target CPI or PCE?
PCE. The Fed's 2% inflation target — locked in the January 2012 FOMC Statement on Longer-Run Goals and Monetary Policy Strategy, and reaffirmed in the August 2020 update — is denominated in personal consumption expenditures price index (PCE), specifically the all-items series. Media often shorthand this as "the Fed targets 2% inflation," which implies CPI in casual usage; the official target is PCE.
Why does PCE run lower than CPI on average?
Three structural differences. (1) Formula: PCE uses a chain-weighted Fisher Ideal index that captures substitution as consumers shift away from items whose prices rise; CPI uses a modified Laspeyres formula with a fixed basket updated every two years, which doesn't capture substitution as quickly. (2) Weighting: CPI gives owner's-equivalent rent roughly twice the weight (~24% vs ~12% in PCE), and shelter inflation has been a major driver of the gap recently. (3) Scope: PCE includes employer-paid health insurance and Medicare/Medicaid spending; CPI is out-of-pocket only.
Which inflation measure should I use for backtesting?
Depends on the question. For Fed-reaction-function models or anything pricing the federal funds rate path, use Core PCE — that's what the Fed reads. For TIPS pricing or COLA adjustments, use CPI by indenture (TIPS reference CPI-U specifically). For broad cost-of-living narratives, CPI is what the media reports. For sector-specific work (housing, health care), the weighting differences make one or the other dominant by topic.
How big is the typical CPI-PCE wedge?
About 0.3 percentage points on average over 2010-2024, with PCE running below CPI. The wedge isn't constant — it widens when housing prices outpace health-care prices (because CPI overweights shelter), and narrows in the opposite regime. In some 12-month windows the wedge has been near zero; in others it's exceeded a full percentage point. Always check the specific window before treating "0.3pp" as a universal constant.
When do CPI and PCE get released each month?
CPI: typically the 10th to 15th business day of each month, for the prior month's data (BLS publishes a release calendar a year in advance). PCE: typically the last Friday of each month, for the prior month's data — PCE incorporates the CPI release that preceded it by roughly two weeks, so PCE always lags CPI within any given month.
Is one more accurate than the other?
Neither is more accurate; they measure different things. CPI measures out-of-pocket household consumer prices for a fixed-weight basket; PCE measures total personal consumption (including employer and government payments) for a chain-weighted basket. Both are correct for their respective questions. The distinction matters most when reading Fed minutes (Core PCE) versus reading TIPS pricing (CPI-U).