TL;DR. The Consumer Price Index (CPI) and the Personal Consumption Expenditures (PCE) price index both measure how fast US prices are rising, but they are built from different data, with different formulas, by different agencies. The Federal Reserve anchors its 2% inflation goal to PCE (specifically core PCE) because PCE covers more of what households actually consume, updates its spending weights more often, and lets people substitute toward cheaper goods the way real shoppers do.
If you have ever wondered why one headline says inflation is 3.0% and another says it is 2.5% in the same week, this is the answer. Both numbers can be right at the same time because they are measuring slightly different things.
Two inflation gauges, two agencies
The Consumer Price Index is produced monthly by the Bureau of Labor Statistics (BLS). CPI tracks the average price change for a representative basket of goods and services bought by urban consumers. It is the index that drives Social Security cost-of-living adjustments, TIPS payouts, and most news headlines.
The PCE price index is produced monthly by the Bureau of Economic Analysis (BEA) inside the Personal Income and Outlays release. PCE measures price change for all goods and services consumed by households and the nonprofits that serve them, sitting inside the broader national-accounts (GDP) framework.
An analogy. PCE is the satellite view, CPI is the street view: same city, different angle and different field of view.
Where the methodologies diverge
Four design choices drive most of the gap between the two indexes: scope, weights, formula, and source data.
Scope. CPI tracks out-of-pocket spending by urban households. PCE is broader: it also captures third-party spending on households’ behalf, including the employer-paid portion of health insurance and the value of medical care delivered through Medicare and Medicaid. That alone is why healthcare has a much bigger footprint inside PCE than inside CPI.
Weights. CPI uses a mostly fixed basket whose category weights are refreshed roughly every one to two years from the Consumer Expenditure Survey. PCE uses chain-weighted weights that are refreshed about each quarter from business-side data inside the National Income and Product Accounts. Because PCE weights move sooner, the index reflects spending shifts faster.
Formula. CPI is a modified Laspeyres-style index that compares today’s prices to a fixed prior-period basket. PCE uses the Fisher-Ideal formula, the geometric mean of a Laspeyres and a Paasche index, which mathematically lets the basket bend toward what consumers are buying now. That bending is what economists mean by “substitution.” If beef gets expensive and households switch to chicken, PCE registers some of the switch; CPI’s fixed basket does not, at least until the next reweight.
Source data. CPI weights come from a household-side survey (what people say they spent). PCE weights come from business records and tax filings inside the national accounts (what stores, hospitals, and providers say they sold). Different vantage points, different totals.
The persistent CPI–PCE wedge
Over long stretches, year-over-year CPI inflation tends to print a few tenths of a percentage point hotter than PCE. The BEA’s own reconciliation work attributes that wedge to a stack of effects rather than one cause, but three usually dominate.
The shelter weight gap. Housing carries a much bigger weight in CPI than in PCE — roughly a third of the headline CPI basket versus closer to one-sixth of PCE. When rents and owners’ equivalent rent run hot, that gap alone can push CPI noticeably above PCE for months.
Substitution bias. A fixed basket cannot let households dodge price spikes by switching to cheaper alternatives. PCE’s Fisher formula can. The result is a small but persistent downward bend in PCE whenever relative prices move.
Scope. Because PCE includes employer-paid health insurance, Medicare, and Medicaid, large parts of medical care enter PCE at administered or negotiated prices that move differently from the retail medical prices CPI surveys.
None of this means one index is wrong. They answer slightly different questions about slightly different populations, and the gap between them is itself a useful signal.
| Feature | CPI (BLS) | PCE (BEA) |
|---|---|---|
| Publisher | Bureau of Labor Statistics (Dept. of Labor) | Bureau of Economic Analysis (Dept. of Commerce) |
| Scope | Out-of-pocket spending by urban households | All spending by & on behalf of households + nonprofits, including employer-paid health insurance, Medicare, Medicaid |
| Weighting method | Modified Laspeyres (mostly fixed basket) | Fisher-Ideal chain index |
| Weight update frequency | Roughly every 1–2 years | Roughly each quarter |
| Approximate shelter weight | ~one-third of the basket | ~one-sixth of the basket |
| Substitution treatment | Limited; consumers stuck with prior basket between reweights | Built into the Fisher formula |
| Source data | Consumer Expenditure Survey (households) | NIPA business records, tax filings, regulatory data |
| Fed’s role | Monitored; underpins TIPS & Social Security COLAs | Primary inflation gauge; the 2% longer-run target is set on PCE |
Why the Fed targets PCE
In its February 2000 Monetary Policy Report to Congress (the “Humphrey-Hawkins Report”), the FOMC said it would begin framing its inflation forecasts in terms of the chain-type PCE price index rather than CPI. The rationale was methodology: PCE’s broader scope, more current weights, and substitution-friendly formula produced what officials viewed as a more consistent, less biased series.
Twelve years later, in January 2012, the Fed went further. The FOMC’s first “Statement on Longer-Run Goals and Monetary Policy Strategy” explicitly defined the 2% inflation target as the annual change in the PCE price index. Every revision of that statement since has kept PCE as the target measure.
Two refinements matter in practice. First, the operational gauge inside the Fed is core PCE, which excludes food and energy. Those categories swing on weather, oil shocks, and geopolitics in ways monetary policy cannot influence, so stripping them yields a cleaner read on trend inflation.
Second, since Chair Powell’s November 2022 Brookings speech, the Fed has spotlighted core services excluding housing — nicknamed “supercore” PCE. Powell called it the largest piece of core PCE and argued that, because labor costs dominate these services, it best reveals how wage pressure feeds into prices. Supercore is not a separate target; it is a diagnostic alongside core PCE.
A simple worked example
Round, illustrative numbers only. Suppose a household’s entire spending is shelter, gasoline, and groceries. Imagine CPI assigns weights of 33% shelter, 10% gasoline, 57% groceries, while PCE assigns 17% shelter, 6% gasoline, 77% groceries. Gasoline prices jump 20% in a month; nothing else moves.
CPI’s headline rises by roughly 0.10 × 20% = 2.0 percentage points. PCE’s rises by roughly 0.06 × 20% = 1.2 points. Same shock, different headline impact, purely from weights.
Now add substitution. Households cut back on driving when gas spikes, so next quarter they buy less. PCE’s chain-weighted formula picks that up quickly and trims the gasoline weight; CPI’s fixed basket holds the old weight until the next reweight. PCE stays muted, CPI stays loud. Stretch that across years and dozens of categories and you have the persistent CPI–PCE wedge.
Common mistakes
- Mixing up headline and core. Headline includes food and energy; core excludes them. Comparing headline CPI to core PCE is apples to oranges, and the gap can be a full percentage point or more.
- Comparing year-over-year CPI to month-over-month annualized PCE. A 0.3% monthly PCE print annualizes to roughly 3.7%, which is not the same as 3.7% YoY. Match the horizons before you compare.
- Treating CPI’s shelter index as real-time. CPI’s shelter component samples rents on a lagged cycle, so it can keep climbing for months after market rents have rolled over.
- Thinking the Fed targets “headline anything.” The Fed’s longer-run goal is 2% headline PCE over time, but the operational gauge it watches month to month is core PCE, supplemented by trimmed-mean and supercore cuts.
- Assuming TIPS are priced off PCE. Treasury Inflation-Protected Securities and most Social Security cost-of-living adjustments are indexed to CPI (CPI-U or CPI-W), not PCE.
- Reading a single month as a trend. Both series are noisy. Three-, six-, and twelve-month annualized changes tell very different stories in any given month.
How to read the data yourself
CPI lands first. BLS releases CPI for the prior month around mid-month, on the schedule posted on its CPI program page. FRED mirrors the headline series as CPIAUCSL and the core as CPILFESL.
PCE lands second. BEA publishes it inside the monthly “Personal Income and Outlays” report, typically in the last week of the month. On FRED, the headline series is PCEPI and the core is PCEPILFE.
A useful workflow: read CPI to see how prices hit budgets in the moment, then wait for PCE later in the month to see how the Fed is most likely to interpret the same story. When the two diverge, look at shelter, healthcare, and substitution-sensitive categories first.
Related concepts to learn next
- Sticky vs flexible CPI (Atlanta Fed) — splits CPI components by how often their prices change.
- Trimmed-mean PCE (Dallas Fed) — a robust core measure that drops the most extreme price movers each month.
- Real interest rates — nominal yields minus expected inflation; the variable monetary policy actually moves.
- Breakevens and TIPS — market-implied CPI inflation expectations from the spread between nominal Treasuries and inflation-protected ones.
- NIPA accounts — the broader BEA framework that produces PCE alongside GDP.
Sources
- BLS — CPI program homepage
- BLS — CPI Handbook of Methods
- BEA — PCE Price Index
- BEA — NIPA Handbook
- BEA FAQ — What accounts for the differences between the PCE price index and the CPI?
- BEA — A Reconciliation between the CPI and the PCE Price Index (paper)
- Federal Reserve — February 2000 Monetary Policy Report (PCE adoption)
- Federal Reserve — January 2012 press release on longer-run goals
- FOMC — Statement on Longer-Run Goals and Monetary Policy Strategy (PDF)
- Federal Reserve FAQ — Why 2 percent inflation?
- Jerome Powell speech, November 30, 2022 (“supercore” services ex housing)
- FRED — CPIAUCSL (headline CPI)
- FRED — CPILFESL (core CPI)
- FRED — PCEPI (headline PCE)
- FRED — PCEPILFE (core PCE)
Disclosure: This article was produced with AI assistance and reviewed before publication. It is for informational purposes only and is not investment advice.