The CPI is not a measurement of any one consumer's inflation experience. It is a weighted average of price changes for a defined basket of goods and services, where the weights reflect average household spending patterns derived from a periodic Consumer Expenditure Survey. The result is a single index number that approximates aggregate consumer-price change while necessarily mis-approximating most individual households.
This page walks through the four design decisions that shape the index: the basket, the weights, the substitution treatment, and the chain-weighting choice. Each decision is defensible and each has critics. Understanding which decisions were made is essential to interpreting the published number sensibly.
1. The basket
The US BLS CPI basket has eight major categories:
| Category | Approx. weight | Examples |
|---|---|---|
| Housing | ~33% | Rent, owner-equivalent rent, utilities |
| Transportation | ~15% | Vehicles, fuel, insurance |
| Food & beverages | ~13% | Groceries, restaurants |
| Medical care | ~8% | Insurance, services, prescriptions |
| Education & communication | ~6% | Tuition, telecoms |
| Recreation | ~6% | Hobbies, entertainment |
| Apparel | ~3% | Clothing, footwear |
| Other | ~16% | Personal care, miscellaneous |
The basket is updated every two years for the BLS CPI-U. The ONS UK CPI updates the basket annually (a publicised event each January where new items are added and obsolete items removed). Both agencies maintain detailed published documentation of the basket composition.
2. The weights
Weights are derived from the Consumer Expenditure Survey (CEX) in the US, the Living Costs and Food Survey (LCF) in the UK. Households complete spending diaries; spending patterns are aggregated and used to weight category price changes in the index.
Two consequences:
- Weights reflect average spending. A household that spends 50% of income on housing experiences a higher inflation rate than the CPI implies during a housing-price spike, because their personal weight on housing is higher than the CPI weight.
- Weights are stable but lag. The 2022–2024 weight set reflects 2019–2021 spending patterns, captured before recent shifts in remote-work, healthcare, and travel spending.
3. The substitution problem
If the price of beef rises sharply, consumers substitute toward chicken. The household's actual cost of getting protein on the table goes up less than the price of beef. The CPI was historically calculated as a Laspeyres-style index that fixes the basket and so overstates the realised inflation experience — this is the classic “substitution bias.”
Modern CPI methodology partially corrects this. The C-CPI-U (Chained CPI for All Urban Consumers) uses a chained Tornqvist formula that allows the basket weights to update each month, capturing realised substitution. The headline CPI-U still uses the older fixed-basket method but applies geomean averaging within categories to capture some lower-level substitution.
4. The chain-weighting controversy
Switching the headline US CPI to chain-weighting (C-CPI-U) instead of CPI-U would lower published inflation by approximately 0.25 percentage points per year — a small annual figure that compounds substantially over decades. The 1996 Boskin Commission recommended exactly this switch; political resistance has prevented its adoption for the headline series, because numerous federal benefits (Social Security, federal tax brackets) are indexed to the headline CPI-U and a switch would slow the growth of those benefits.
5. What the headline figure actually measures
The published CPI-U is best understood as: the answer to “how much would it cost today to buy approximately the same basket of goods and services as the average urban consumer bought a year ago.” The basket is fixed at the reference period (with periodic updates); substitution is partly handled within categories; quality changes are adjusted via hedonic methods (see the hedonic page).
The figure is a defensible aggregate measure. It is not a forecast of personal experience.