Published by Instant Press on behalf of Fight For A Living Wage.
Inflation gets measured by pricing a fixed basket of goods and services, then watching what that basket costs over time. The Bureau of Labor Statistics sends collectors to stores and service providers every month, records thousands of individual prices, and weights each category by how much a typical urban household spends on it. The weighted total becomes the Consumer Price Index. Almost every argument about whether inflation figures are accurate is an argument about the weights.
How the basket gets built
The spending weights come from the Consumer Expenditure Survey, in which households record what they bought. If housing accounts for roughly a third of typical spending, housing carries roughly a third of the index. Food, transportation, medical care, education, recreation and apparel each get a share proportional to observed spending.
Collectors then price specific items, not categories. Not bread but a named loaf at a named store in a named city. Holding the item constant is what makes month-to-month comparison meaningful, and it is also the source of most of the methodological difficulty, because real products change.
The BLS documents its procedures, publishes the category detail underneath the headline figure, and releases regional and metro-level indexes. All of it sits at bls.gov.
Two adjustments that cause arguments
Substitution comes first. When the price of beef rises and shoppers buy more chicken, a strictly fixed basket overstates the cost of maintaining a standard of living, because it charges households for a purchase they stopped making. Allowing for substitution corrects that. It also means the index measures the cost of an adjusted standard of living rather than an identical one, which is a real conceptual shift.
Quality adjustment, or hedonic adjustment, comes second. A laptop costing the same as one from five years ago is a faster machine, so part of an unchanged price represents more value received. Statisticians strip that out to isolate pure price change. The procedure is defensible and the magnitude is contestable, and it cuts hardest in categories like electronics and vehicles where features change fastest.
Housing gets its own treatment. Most of the shelter component is owners’ equivalent rent, an estimate of what an owner-occupied home would rent for, because a house purchase is an investment as well as a consumption good. That estimate moves more slowly than market rents and more slowly than home prices, which is why the shelter line can look calm during a period when buying a house clearly got harder.
CPI and PCE are not the same thing
Two inflation measures circulate and they answer different questions. The CPI tracks out-of-pocket spending by urban consumers. The personal consumption expenditures price index, published by the Bureau of Economic Analysis, covers a broader set of purchases including medical care paid by employers and government on a household’s behalf, and it updates its weights more frequently.
PCE usually runs a little below CPI, mostly because of those weighting and coverage differences. The Federal Reserve targets PCE. Cost-of-living adjustments in federal programs generally use a CPI variant. Reading one number against a rule written for the other is a common way to reach a wrong conclusion. Federal Reserve materials on the target and the data are at federalreserve.gov.
Headline and core add another split. Core strips out food and energy, which swing on weather and global supply and can drown out the underlying trend. Core is the better predictor and the worse description of a household’s month, since food and energy are the bills people notice first.
Why the index misses your year
The weights describe a typical household. Your household is specific, and the mismatch is where the measure stops matching experience.
Three categories make the point. KFF put total annual premiums for employer family health coverage near $25,000 in 2024, with the worker’s share above $6,000. Child Care Aware reports center-based childcare commonly at $10,000 to $17,000 or more per child per year. National Association of Realtors and Census figures put the median U.S. home sale price at roughly $400,000 to $420,000 in 2024, close to five times the median household income of about $80,000 the Census Bureau reported for 2023, against roughly three times in the 1980s.
A family renting, paying for daycare and carrying an employer plan is concentrated in those three lines. A household that owns outright, has Medicare and has no children at home has nearly no exposure to any of them. The index reports one number to both. It is accurate about the aggregate and uninformative about either case, and the household with heavy exposure is right when it says the figure understates its year.
What a fixed nominal amount does over time
The clearest use of an inflation measure is converting a past dollar figure into current purchasing power. The federal minimum wage has been $7.25 an hour since 2009, according to the U.S. Department of Labor. Applying the published CPI series to that unchanged figure shows what has happened to its real value across the intervening years, and the BLS inflation calculator does the arithmetic directly.
Fight For A Living Wage, a nonpartisan 501(c)(3) grassroots organization, treats affordability across housing, health care, childcare, food, transportation and education as the central issue rather than the wage floor on its own. The category detail in the CPI release is what makes that distinction measurable, since it shows which parts of the basket moved and which did not.
Using the number well
The index is a good instrument used as designed: a consistent national series for comparing one period against another. It fails when it gets asked to describe a particular household, because the weights that make it representative are the same weights that make it wrong for anyone specific. The fix is available in the same release. Open the category tables, find the lines that match your own spending, and read those instead of the headline.
