Macroeconomics › Section I · Inflation › Topic I.2
Measuring changes in the cost of living
The Consumer Price Index is the number behind pay rises, pension increases and tax brackets, so how it is built matters to almost everyone. This topic opens the basket, shows why a fixed basket tends to overstate the rise in living costs, and compares the CPI with the other price indexes economists use.
Who builds the CPI, and how
In the United States the CPI is produced monthly by the Bureau of Labor Statistics (BLS). Its main version, the CPI-U, covers all urban consumers, who make up over 90% of the population; a narrower version, the CPI-W, covers households of wage earners and clerical workers, about 30% of the population. BLS staff record the prices of about 80,000 items each month and survey rents separately. The weights come from the Consumer Expenditure Surveys of household spending. Since 2023 the BLS has refreshed those weights every year, using spending from a single recent year.
What is in the basket
The BLS groups spending into eight major categories. According to its relative importance figures for December 2025, the weights were roughly: housing 44.5% (of which shelter, mainly rent and the rental value of owned homes, was 35.6%), transportation 16.3%, food and beverages 14.5%, medical care 8.4%, education and communication 5.8%, recreation 5.1%, other goods and services 2.9% and apparel 2.4%. The weights explain why some price changes move the CPI so much more than others. A 5% rise in shelter costs adds about 1.8 percentage points to the index; a 5% rise in clothing adds about 0.1.
A fixed basket versus the cost of living
Ideally we want to know how much more money a household needs to stay as well off as before. That is a cost-of-living index. A fixed-basket index answers a narrower question: how much more does the same list of goods cost? The two differ because people switch between products, get better versions of old ones, and buy things that did not exist when the basket was chosen. The BLS itself says the CPI only approximates a cost-of-living index. The gaps between the two are usually called biases.
Substitution bias
When one good becomes relatively dearer, households buy less of it and more of its substitutes. A fixed basket ignores this, assuming people go on buying the same quantity of the dearer good. It therefore tends to overstate the rise in the cost of staying equally well off. Substitution happens within a category, such as switching coffee brands, and across categories, such as eating at home more when restaurant prices jump. The bias is largest when relative prices move sharply.
Worked example: switching from beef to chicken
A family buys 10 pounds of beef at $5 and 10 pounds of chicken at $3 each month. 1. Original cost: \(50 + 30 = 80\) dollars. 2. Beef rises to $7; chicken stays at $3. 3. Fixed-basket cost: \(70 + 30 = 100\) dollars, a rise of \(20 \div 80 = 25\%\). 4. Suppose the family is about as happy with 6 pounds of beef and 14 pounds of chicken. That costs \(42 + 42 = 84\) dollars, a rise of only \(4 \div 80 = 5\%\). 5. If the family really is as well off with the new mix, its cost of living rose by about 5% at most, so the fixed basket's 25% overstates the rise; the poorer a substitute chicken is, the smaller that gap.
Quality, new goods and new outlets
Quality bias arises when a price rises partly because the product got better. If a phone costs 10% more but has a far better camera and battery, counting the whole 10% as inflation overstates it. New goods bias arises because new products enter the basket only after a delay, often after their prices have already fallen steeply, so the index misses that early fall and the gain from having the product at all. Outlet bias arises when shoppers move to cheaper sellers, such as discount stores or online retailers, and the index keeps sampling the dearer ones. Each of these pushes the measured CPI above the true change in living costs.
How the BLS reduces the biases
Since 1999 the BLS has combined prices within most item categories using a geometric mean, a formula that allows for some substitution. Since 2002 it has also published the chained CPI-U, which allows for substitution across categories but has to be revised for months afterwards. For some products, such as many clothing items, it uses hedonic models that estimate how much of a price change reflects changed features. Yearly weight updates, rotating samples of stores and items, and price quotes collected from retailers' websites also help. In 1996 the Boskin Commission estimated an upward bias of about 1.1 percentage points a year; later estimates, after these changes, are generally smaller but still positive.
Core inflation
Food and energy prices swing with weather, harvests, wars and refinery outages, and those swings often reverse within months. Core inflation strips them out, using the index for all items less food and energy, to show the underlying trend. Core is not what households pay, since families still buy food and fuel, but it is a better guide to where inflation is heading, which is why central banks watch it. When headline runs well above core, food or energy is the reason.
Other price indexes
The CPI is not the only gauge. The personal consumption expenditures (PCE) price index, produced by the Bureau of Economic Analysis (BEA), covers spending by households and on their behalf, such as employer-paid health insurance, and its formula allows for substitution. The Federal Reserve defines its 2% inflation goal in terms of the PCE index. The Producer Price Index (PPI) measures prices that domestic producers receive, including for inputs. The GDP deflator covers everything in GDP, as the topic "Adjusting nominal values to real values" explains. The Employment Cost Index tracks the cost of pay and benefits, and import and export price indexes track traded goods.
Worked example: reading four measures at once
Official figures for September 2026 showed, over 12 months: CPI 3.7%, core CPI 2.6% (BLS); PCE 3.6%, core PCE 3.1% (BEA). 1. Headline CPI is 1.1 points above core CPI, so food or energy is pulling it up. 2. The BLS release confirms it: energy prices were up 20.4%, with petrol (gasoline) up 34.2%, while food was up only 2.5%. 3. The CPI and PCE disagree by a few tenths because they weight items differently and cover different spending. 4. Against the Fed's 2% PCE goal, both headline and core PCE were above target, though the core gap was smaller.
Common mistakes to avoid
First, thinking the CPI measures your own cost of living: it reflects an average urban household, and a renter or a long-distance commuter may face a different mix. Second, assuming the fixed-basket biases understate inflation: they push the measured rise up, not down. Third, treating core inflation as what households really pay. Fourth, mixing up sources: the CPI comes from the BLS, the PCE index and GDP deflator from the BEA. Fifth, forgetting that the Fed's 2% goal refers to PCE inflation.
Key takeaways
The BLS builds the CPI from about 80,000 monthly price quotes, weighted by household spending, with housing the largest category. A fixed basket overstates the rise in the cost of living because of substitution, quality and new goods, and outlet bias. Geometric means, the chained CPI-U, hedonic adjustment and yearly weight updates shrink these biases without removing them. Core inflation excludes food and energy to show the trend. The PCE index, the Fed's preferred gauge, the PPI, the GDP deflator and the Employment Cost Index each answer a different question.
Flashcards
Flashcard
CPI for all urban consumers, over 90% of the US population
Test your understanding
Question 1 of 5
A household buys 4 cinema tickets at $10 and 4 streaming rentals at $5 a month. Tickets rise to $15. If the basket stays fixed, by how much does its cost rise?
Question 2 of 5
In the same household, why might the fixed-basket figure overstate how much worse off the family really is?
Question 3 of 5
Headline CPI inflation is 3.7% while core CPI inflation is 2.6%. What is the most likely explanation?
Question 4 of 5
Which statement about the PCE price index is correct?
Question 5 of 5
A laptop's price rises 8% this year, but it has a much faster processor and longer battery life. How should a well-built price index treat the 8%?
