The Mechanics of How Inflation Is Measured—And Where the Gaps Are

Inflation is one of the most discussed economic variables in public discourse, cited in central bank decisions, wage negotiations, benefit adjustments, and political arguments. Yet the number reported in headlines as “the inflation rate” is a statistical construct with specific design choices, measurement conventions, and known limitations built in. Understanding what the Consumer Price Index (CPI) and related measures actually track—and where they diverge from economic experience—helps explain why inflation feels different to different people and why economists sometimes disagree about whether reported inflation accurately reflects the price environment.

How the CPI Is Constructed

The Consumer Price Index measures the average change over time in the prices paid by urban consumers for a fixed “basket” of goods and services. The basket is meant to represent the typical spending patterns of households, and the prices in the basket are tracked over time to calculate how much more (or less) it costs to purchase the same set of goods.

In the United States, the Bureau of Labor Statistics (BLS) collects approximately 80,000 price quotes per month from thousands of retail establishments, service providers, rental properties, and other sources. These prices are organised into categories: food and beverages, housing, apparel, transportation, medical care, recreation, education, and other goods and services. Each category receives a weight in the index proportional to how much of household expenditure it represents—housing and transportation together make up about half of the CPI, food about 14%, and medical care about 7%.

The weights are derived from the Consumer Expenditure Survey, which tracks the actual spending patterns of thousands of households. These weights are updated periodically—in the US, they were updated annually starting in 2022, replacing the prior practice of updating every two years—to reflect changes in what people buy. The basket is not entirely fixed; it evolves as spending patterns change.

The Shelter Problem

Housing is the largest single component of the CPI, at roughly a third of the total index weight, and it is measured in a way that creates a significant lag relative to actual market conditions. The BLS primarily measures shelter costs through “owners’ equivalent rent” (OER)—a concept that asks what homeowners would pay if they were renting their own homes. This is measured by surveying homeowners about the hypothetical rent for their unit.

OER does not measure home purchase prices or mortgage rates directly. Instead, it tracks the imputed rental value of owner-occupied housing, which changes slowly as the sample of surveyed homeowners gradually reflects market conditions. Because most homeowners have fixed-rate mortgages and stable housing costs, their responses to the OER survey change slowly even when market rents are rising rapidly.

This creates a structural lag: CPI shelter inflation significantly lags actual market rent inflation, sometimes by twelve to eighteen months. During 2021-2022, when market rents were rising at double-digit annual rates in many US cities, the CPI shelter component was rising much more slowly. Conversely, when market rents began declining in 2023, the CPI shelter component continued rising for many months afterward. This lag means the CPI headline number can diverge substantially from the actual price environment people experience—overstating inflation when rents are falling, understating it when rents are rising quickly.

Quality Adjustment and Hedonic Methods

A fundamental challenge in price measurement is distinguishing changes in price from changes in quality. If a laptop costs 10% more this year than last year but has a processor twice as fast and double the storage, is that inflation or an improvement? The standard answer in price statistics is that some of the price change reflects quality improvement and should not be counted as inflation—which requires making explicit judgments about how to value quality changes.

The BLS uses hedonic regression methods for product categories where quality changes rapidly, particularly electronics and computers. Hedonic models estimate the market value of specific product attributes (processor speed, memory, screen resolution) using price data, then use those estimates to adjust price changes for quality improvements. The result is that measured CPI for electronics tends to fall over time even when nominal prices are stable, because quality improvements are treated as equivalent to price reductions.

Critics argue that hedonic adjustments understate inflation by attributing too much of price changes to quality improvements. Defenders argue the adjustments correctly capture the improving value consumers receive. The debate is genuinely unresolved—reasonable statistical methods can produce meaningfully different estimates. What’s clear is that the choice of hedonic methodology has a measurable effect on reported inflation, and the effect is not trivial for categories like healthcare and technology where quality changes are large.

The Substitution Bias Question

The CPI uses a “fixed basket” conceptually, but with a specific chain-linking modification. A pure fixed basket index would measure the cost of exactly the same goods each period, which would overstate inflation because it ignores that consumers substitute cheaper alternatives when prices rise. If beef prices rise, some consumers buy more chicken; a fixed-basket index that keeps counting beef purchases ignores this adaptation.

The BLS uses a “chained CPI” (C-CPI-U) that accounts for substitution by linking expenditure patterns across periods. The chained CPI consistently runs about 0.25-0.30 percentage points lower than the standard CPI-U annually, over time accumulating to significant differences in cost-of-living adjustments. The personal consumption expenditure (PCE) deflator, which the Federal Reserve uses as its primary inflation measure, also accounts for substitution and typically runs slightly lower than the CPI-U.

Whether substitution adjustments are appropriate depends on the purpose. If you want to know how much more it costs to maintain the same standard of living, adjusting for substitution makes sense. If you want to know how much prices have risen for a specific bundle of goods (for, say, a cost-of-living adjustment for a person who can’t or won’t substitute), the fixed basket measure is more appropriate. The different inflation measures represent different answers to slightly different questions—which is why seemingly straightforward questions about “the inflation rate” can get different answers depending on which measure is used.

What CPI Misses

The CPI measures a specific concept—the average price change for an average urban consumer—and by design doesn’t measure several things people might care about. Asset prices (housing values, stocks) are not in the CPI; it measures the cost of consuming housing (rent or OER) but not the price of owning housing as an asset. During periods of rapid asset price appreciation, this exclusion means the CPI understates the cost of achieving a given level of wealth accumulation even if it accurately tracks consumption costs.

The CPI is designed around average households. Households with different spending patterns—the elderly, low-income households, people in high-cost cities—experience different inflation rates than the headline CPI suggests. The BLS produces an experimental CPI-E for the elderly, which historically runs slightly higher than the standard CPI because the elderly spend more on healthcare (which has inflated faster than average) and less on education and technology (which have inflated less). The degree to which the CPI measures the inflation experienced by any specific household depends on how similar that household’s spending is to the national average.

The CPI also struggles with new goods: a new product category isn’t in the basket until it’s added, which means the CPI misses deflationary effects from entirely new product categories (free internet services, for example, aren’t in the basket but represent value that wasn’t previously available). The timing and methodology for adding new categories to the basket create a structural lag that may systematically understate improvements in living standards even when they’re accurately measuring prices of existing goods.

None of these limitations make the CPI a poor measure—it is carefully constructed, methodologically transparent, and serves its primary purposes well. But treating it as a perfect representation of “the cost of living” misses the significant technical choices and known gaps that any serious interpretation of inflation data requires acknowledging.