Two measures can disagree without either being wrong
An inflation rate is a change in a defined price index, not a direct reading of every household’s financial strain. The Consumer Price Index for All Urban Consumers, or CPI-U, follows consumer prices for a specified urban population. The Personal Consumption Expenditures price index, or PCE, measures prices within the national accounts’ broader personal-consumption boundary. Comparing them requires the same time period and an understanding of what each includes. [1]
The distinction between a price level and its growth rate is equally important. In a hypothetical sequence, an index rises from 100 to 103 and then to 105.06. Inflation slows from 3% to 2%, but the basket is still 5.06% more expensive than at the start. Slower inflation does not restore the old price level. Whether purchasing power improved also depends on income, taxes and the household’s particular purchases.
Who pays changes the boundary
BEA separates CPI–PCE differences into formula, weight, scope and other effects. PCE covers spending by and on behalf of the personal sector, including households and nonprofit institutions serving them. CPI’s expenditure scope is principally household out-of-pocket spending. Thus a health service paid for by an employer or government program can matter differently from the same service paid directly by a patient. This is a coverage distinction, not evidence that one agency thinks medical treatment is unimportant. Seasonal adjustments and underlying price information can create additional differences. [1]
The expenditure boundary is different from the price observed for a medical service. BLS explains that medical-service price changes include total reimbursement, including insurer payments and patient copayments, even though the overall medical-care expenditure weight is based on the CPI’s narrower out-of-pocket scope. An unchanged copayment therefore does not by itself imply unchanged CPI medical inflation. The scope difference concerns which spending is represented and how heavily it is weighted, not a blanket exclusion of insurer payments from every service-price observation. [2]
The same prices, different weights
The following deliberately simplified example isolates weighting. Assume three categories have price increases of 6%, 2% and 1%. Basket A allocates more spending to shelter than Basket B. Applying fixed starting shares gives 2.95% inflation in A and 2.30% in B: a 0.65-percentage-point gap with no disagreement about any individual price. These are invented baskets, not actual CPI or PCE weights or an exact reconstruction of either index.
The arithmetic also explains why the gap can reverse. If shelter inflation instead falls to 0% while the other two categories keep their assumed increases, A rises 0.85% and B rises 1.10%. A heavier shelter weight now lowers measured inflation. The direction of an index gap depends on relative price changes as well as weights; a historical average difference is not a permanent conversion factor.
Scroll horizontally to see all columns.
| Hypothetical category | Price change | Basket A weight | Basket B weight |
|---|---|---|---|
| Shelter | 6% | 35% | 20% |
| Health services | 2% | 20% | 30% |
| Other consumption | 1% | 45% | 50% |
| Weighted result | Not a category | 2.95% | 2.30% |
Changing baskets is more subtle than a fixed-versus-flexible slogan
PCE uses a Fisher-Ideal formula, while CPI-U uses modified Laspeyres aggregation. [1] That does not mean CPI prices an unchanged shopping list forever. BLS uses both geometric-mean and modified-Laspeyres formulas for detailed indexes and updates CPI-U expenditure weights annually. The separate chained CPI is another index, with different aggregation and revision characteristics; it is not simply a second name for PCE. [4]
Conceptually, substitution creates two legitimate questions: how much more does an unchanged basket cost, and how much more spending is needed when people can alter their purchases? If a household switches from an expensive meal to a cheaper substitute, its spending may rise less than the original menu’s cost. That does not establish that it feels equally well off. The interpretive disagreement concerns the standard of comparison, not only arithmetic.
Product replacement adds another layer. BLS adjusts for changes in quality when old items disappear or new versions replace them, seeking comparable consumption rather than counting every specification change as inflation. The adjustment is an estimate, not a claim that shoppers receive cash back for a better feature. [4] A larger purchase price can therefore coexist with a smaller measured constant-quality price increase.
Housing measures a service, not the purchase of an asset
For owner-occupied housing, CPI estimates the change in the rental value of shelter through owners’ equivalent rent. It does not insert house sale prices or mortgage payments into that component. BLS uses rental observations to measure price changes; homeowners’ answers about possible rent help determine expenditure weights, not monthly price movements. Mortgage interest and house purchases fall outside this shelter-consumption measure. [3]
This explains an apparent paradox: someone with an unchanged fixed mortgage payment can live in an area with rising measured shelter costs. The index is valuing housing services, while the homeowner is describing a financing contract. A new buyer’s affordability problem can also worsen because of financing costs even when shelter inflation is moderating.
BLS collects each sampled rental unit’s rent every six months, with different panels observed in different months. Many sampled leases are continuing contracts. Consequently, a series focused on newly advertised leases and CPI rent need not turn at the same time. Differences in the population of rents and collection timing matter; one is not automatically evidence that the other is defective. [3]
Monthly momentum and year-over-year inflation answer different questions
Seasonal adjustment attempts to separate recurring calendar patterns from other changes. BLS recalculates seasonally adjusted CPI indexes for the previous five years as it updates seasonal factors. [4] That makes the adjustment status and data part of a meaningful comparison, rather than minor details of a chart label.
Hypothetically, a monthly increase of 0.2% compounds to approximately 2.43% over twelve months if repeated; 0.4% compounds to approximately 4.91%. Those annualized calculations are scenarios, not forecasts. A twelve-month rate instead compares the current index with its actual value twelve months earlier. It can fall when an unusually large old monthly increase drops out even if the latest monthly pace is unchanged.
For a household example, assume a rent renewal absorbs a much larger share of spending than in either national index. The same rent increase can then create a larger personal-budget shock even if every quoted category price is measured accurately. Individual affordability also depends on income and available substitutes. A national aggregate and that household’s experience can both be internally consistent.
What the disagreement can reveal
An explanatory comparison separates a broad change in price pressure from a shift concentrated in categories that one index weights more heavily. Agreement across the measures can strengthen the interpretation of a widespread trend, but it does not establish the distribution of that trend across households. Disagreement that can be traced to coverage or shelter weights can be economically coherent.
The evidence that clarifies the gap is the agencies’ category detail, expenditure weights, common-period changes and reconciliation tables. A claim that one index is universally superior needs a specified purpose: an indexed contract, a household-budget question and a national-consumption question need not use the same benchmark. This article explains those differences; it does not estimate a live inflation reading or prescribe a financial response.
Sources
- BEA; What accounts for the differences in the PCE price index and the Consumer Price Index?; November 3, 2010, current page checked October 4, 2026Official sourceBack to text: ↑1↑2↑3
- BLS; Measuring Price Change in the CPI: Medical care; current methodology checked October 4, 2026Official sourceBack to text: ↑
- BLS; Measuring Price Change in the CPI: Rent and Rental Equivalence; current methodology checked October 4, 2026Official sourceBack to text: ↑1↑2↑3
- BLS; Consumer Price Index Handbook of Methods, Calculation; current methodology checked October 4, 2026Official sourceBack to text: ↑1↑2↑3