Akshay DineshA grocery receipt looks ordinary: milk, bread, detergent, perhaps a last-minute snack. Yet when thousands of households record purchases like these, the entries become part of one of the most useful pictures of American economic life. The U.S. Bureau of Labor Statistics uses the Consumer Expenditure Surveys to estimate how people divide their money among housing, food, transportation, health care, entertainment, and hundreds of more specific categories.
The numbers often appear simple. In the 2025 annual release, for example, the average consumer unit spent $81,905, up 4.3 percent from 2024. But that result did not come from adding up cash-register totals or asking one broad question about a family budget. It came from two different surveys, carefully selected addresses, repeated interviews, daily records, statistical weights, and many decisions about what counts as an expenditure. Knowing that machinery makes the final tables far easier to read honestly.
One survey cannot remember every kind of purchase
Think about the last three months of spending. A person can probably recall the monthly rent or mortgage, a car payment, and the purchase of a refrigerator. Remembering every cup of coffee, bus fare, prescription, and bag of groceries is much harder. The Consumer Expenditure program handles that mismatch with two separate data collections, each designed around the way memory works.
The Quarterly Interview Survey focuses on large or recurring expenses that people can usually recall over a longer period. A participating consumer unit is interviewed every three months for four consecutive quarters. The questions cover costs such as housing, utilities, vehicles, insurance, medical care, and major household equipment. Repeated interviews help capture changes across a year instead of treating one month as though it represents every season.
The Diary Survey works at a much shorter range. Participants record purchases for two consecutive one-week periods, including small and frequent expenses that would fade quickly from memory. Detailed food purchases, personal-care products, gasoline, and many other everyday items are better suited to a diary than a three-month recall. The diary asks for all eligible expenses during the week, even though analysts later use it mainly where it produces the stronger measurement.
The two samples are separate, and their coverage overlaps. BLS does not simply count the same purchase twice. For each expenditure category in the integrated annual tables, analysts choose the survey that measures it more reliably. Rent is naturally suited to an interview; detailed groceries are more naturally suited to a diary. The result is a statistical patchwork, but it is a deliberate one: each piece comes from the instrument best able to capture it.
The survey measures consumer units, not just households
The basic unit in the data is called a consumer unit. It may be a family related by birth, marriage, adoption, or another legal arrangement; a person living alone; or unrelated people who live together and pool income to make major spending decisions. Two roommates who keep their finances separate may belong to different consumer units even though they share one address. Students in university housing can also be counted as separate units.
That distinction matters whenever a table reports spending “per household.” The convenient everyday word may hide a more precise statistical definition. A consumer unit is built around shared financial decisions, not merely a roof. It gives analysts a better way to connect purchases with income, family composition, age, employment, and other characteristics.
An expenditure also has a specific meaning. BLS generally records the financial obligation for a good or service when it is acquired, including applicable sales and excise taxes. Reimbursed and business-related costs are excluded. The survey is not a complete measure of everything people consume: a gift received from someone outside the unit, food produced at home, barter, and many in-kind benefits can be consumed without becoming a consumer expenditure.
This boundary explains why Consumer Expenditure totals do not perfectly match other familiar measures. The Bureau of Economic Analysis builds Personal Consumption Expenditures largely from business and administrative records and uses somewhat different concepts. A family budget, a Consumer Expenditure estimate, and national PCE data can all describe spending while answering different questions.
How a small sample becomes a national estimate
Survey respondents represent far more than themselves. The Census Bureau selects addresses for both surveys and collects the responses on behalf of BLS. Each usable consumer unit receives a statistical weight based first on its probability of selection. BLS says a typical base weight is around 10,000, meaning one responding unit may initially stand for roughly 10,000 units in the civilian noninstitutional population.
The base weight is only the beginning. Analysts adjust for nonresponse, because people who complete a demanding spending survey may differ from those who do not. They also calibrate the sample so its demographic totals better match independent population information. Reported purchases are checked for consistency, coded into common categories, corrected when they are clearly misclassified, and sometimes imputed when a usable value is missing.
Time has to be standardized too. Interview responses cover three-month recall periods, while diary entries cover a week. Annual diary estimates are scaled from weekly records, and interview data are aligned with the calendar year. Once the data are prepared, weighted expenditures are combined to estimate an average annual amount per consumer unit.
This is why the published average is not the same as the sum of receipts divided by the number of respondents. Some participants represent more of the population than others, and each survey observes spending over a different window. The method is trying to reconstruct the nation, not merely summarize the people who happened to answer.
An average is a map, not a model budget
The annual tables include all consumer units in a group when calculating the average for a category, including units that reported no spending on that item during the period. If only a minority buys a new vehicle, the average vehicle-purchase figure is spread across buyers and nonbuyers alike. It may therefore be far lower than the amount paid by people who actually made a purchase.
The mean can also sit well above or below what many individual families spend. Household size, income, age, region, housing tenure, health needs, and access to transportation all reshape a budget. A national average is useful for comparing groups and watching broad changes, but it is not a prescription for how much one family should spend.
Even year-to-year changes need context. The 2025 release reported that average expenditures rose 4.3 percent while the Consumer Price Index for All Urban Consumers rose 2.6 percent. Those percentages are not directly interchangeable. Spending can rise because prices increased, because people bought different quantities or qualities, because the population mix changed, or because several forces moved at once.
The 2025 data also carry an unusual methodological wrinkle. A federal lapse in appropriations stopped Census Bureau collection in October and November 2025. Later interviews recovered some spending through their three-month recall periods, but the interruption still left fewer reports for several months, while the Diary Survey had no collection for October and November. BLS used weight adjustments, including using adjacent diary months to represent the missing period. That does not make the release useless, but it is a real caution attached to comparisons involving 2025.
Why these spending records shape the CPI
The Consumer Expenditure Surveys do more than describe budgets. Their data help determine the relative importance of goods and services in the Consumer Price Index. Prices for housing, food, medical care, transportation, and other categories do not all receive equal weight. The index needs evidence about how spending is distributed before it can combine many price movements into one overall measure.
Imagine that rent rises sharply while the price of pencils falls. A price index should not treat those changes as equally important to a typical budget. Consumer Expenditure data help establish the spending shares that give heavily purchased categories more influence. The survey does not collect the retail prices used in the CPI; it helps answer a different question: how much weight should each area of spending carry?
The data also support research on living costs, retirement, poverty, taxes, nutrition, and the way spending differs across groups. Public-use files let researchers move beyond national averages and examine patterns by age, income, family type, and other characteristics, while confidentiality protections limit the detail that could identify respondents.
A spending table becomes most informative when read with three questions in mind: Who is represented? Which survey captured the category? What does the reported average include? Those questions turn a neat column of dollar amounts back into what it really is: an estimate built from many imperfect human memories and records, strengthened by careful survey design. The receipts may be ordinary, but the measurement behind them is anything but simple.
