When a headline says inflation was 3.2% last month, it is reporting one specific statistic produced by one federal agency using a specific method. Knowing how that number is built explains most of the arguments people have about it — including the very common experience of reading the figure and thinking it feels wrong.
The basic idea: a fixed basket, repriced
The Consumer Price Index measures the average change over time in the prices paid by urban consumers for a representative basket of goods and services. The Bureau of Labor Statistics defines the basket, sends data collectors out to record actual prices each month, and computes how much the cost of that same basket changed.
The basket is not a guess. Its composition comes from the Consumer Expenditure Survey, in which thousands of households record what they actually spend money on. Those spending patterns become the weights: if households spend roughly a third of their budget on housing, housing carries roughly a third of the weight in the index.
That weighting is why not every price increase moves the index equally. A 10% jump in the price of eggs and a 10% jump in rent are not remotely the same event as far as the CPI is concerned.
The major components
The index is organized into eight major groups: housing, food and beverages, transportation, medical care, apparel, recreation, education and communication, and a catch-all "other goods and services."
Housing dominates, and the way it is measured surprises people. Because a house is an asset rather than something consumed in a month, the BLS does not put home purchase prices in the index. It uses owners' equivalent rent instead — essentially, what a homeowner would have to pay to rent an equivalent home. That is a survey-based estimate, it moves slowly, and it lags actual market rents, all of which contributes to the sense that the official number is behind reality.
The versions, and which one touches your money
There is not one CPI. There are several, and different ones are wired into different parts of American life.
CPI-U covers all urban consumers, roughly 93% of the population, and is the headline figure quoted in the news.
CPI-W covers urban wage earners and clerical workers only — a narrower, more employment-weighted population. This is the one the Social Security Administration uses to set the annual cost-of-living adjustment. For 2026, that COLA was 2.8%.
Chained CPI (C-CPI-U) adjusts for substitution: when beef gets expensive and consumers buy more chicken, chained CPI reflects that shift, while the fixed-basket versions largely do not. Because substitution dampens measured increases, chained CPI generally rises a bit more slowly. Since the 2017 tax law, the IRS uses it to index federal tax brackets and the standard deduction — which means brackets creep upward slightly slower than they would have under CPI-U.
That last point has a real consequence: over many years, indexing to a slower-growing measure gradually pulls more income into higher brackets than the older method would have.
Core versus headline
Core CPI strips out food and energy. This is routinely misread as the government hiding the prices that matter, which gets the purpose backwards.
Food and energy prices swing violently for reasons unrelated to broad monetary conditions — weather, geopolitics, a refinery outage. Including them makes month-to-month readings noisy and can point policy in the wrong direction. Central bankers watch core to see the persistent trend underneath the noise.
Both numbers are published every month, neither is hidden, and headline CPI is the one used for the indexing decisions that actually move money.
Why it never matches your experience
The CPI describes an average household, and no actual household is average. Your personal inflation rate depends on the specific basket you buy.
A homeowner ten years into a fixed-rate mortgage has a housing cost that is literally flat in nominal terms, while a renter in a tight market may face double-digit increases — and housing is the largest component of the index for both of them. Add a long commute, or childcare, or a chronic medical condition, and your lived rate can diverge from the national number by several points in either direction.
This is not a flaw in the statistic. It is what an average is. The CPI is the right tool for indexing benefits and comparing decades; it is the wrong tool for predicting your own household budget.
What to do with the number
Use it for what it is good at: converting dollar amounts across time. Asking what a $30,000 salary in 1995 is worth today, or what your savings will buy in twenty years, is exactly the question the CPI answers well.
Do not use a single month's reading to make long-term decisions. Monthly figures are noisy and get revised, and seasonal adjustment can make a given month look better or worse than the trend. Year-over-year comparisons are far more stable.