Why census numbers don't match, and which one to use

Look up how many people live in the United States and you can get several answers from the same agency. The 2020 census counted 331,449,281. The U.S. Census Bureau's population estimates put the total at 341,784,857 on July 1, 2025. Its population clock, on the day you check, shows a third number. None of them is wrong. They come from different programs built for different jobs.

Knowing which one to use is the difference between a number that holds up and one that gets picked apart. Here is how the three main sources work.

The decennial census is the count. It happens every 10 years and tries to reach every person living in the 50 states, the District of Columbia, and the U.S. territories. It provides the official count that decides how many seats each state gets in the House. It asks only a few questions, such as age, sex, race, Hispanic origin, and whether a home is owned or rented.

The American Community Survey is a sample. Every month, the bureau sends it to a share of about 3.5 million addresses a year and asks about topics the census skips, such as income, education, jobs, commuting, and internet access. Because it is a survey, every figure comes with a margin of error. Results come in two forms. The 1-year estimates cover places with at least 65,000 people and are the most current. The 5-year estimates pool 60 months of answers and cover every place down to small towns and census tracts. The bureau says "Only compare 1-year estimates with 1-year estimates, and 5-year estimates with 5-year estimates."

The Population Estimates Program fills the years between censuses. It starts from the last census and adds births, subtracts deaths, and accounts for people moving in and out, producing July 1 estimates for the nation, states, counties, cities, and towns. The population clock is a short-term projection that runs forward from those estimates.

The bureau is specific about which to use. "If you are looking for population totals, we recommend the 2020 Census or Population Estimates Program," it says, and it advises using the survey "to obtain population characteristics (percents, means, medians, and rates) rather than estimates of population totals." In practice, that means using the population estimates for how many people live in a place, and the American Community Survey for what those people are like.

The differences can be large for smaller groups. The population estimates count 13.95 million people who are American Indian or Alaska Native alone or in combination with another race. The 2025 American Community Survey puts the same group at 10.79 million. Both are official, but they answer the question in different ways, and the bureau leans on the estimates for totals.

For the country as a whole, the two line up closely. The survey's national total for 2025 is 341,784,859, two people more than the population estimate. The bigger gaps show up in small places and small groups, where the survey's margins of error widen.

Two other tools mix sources. QuickFacts combines figures from several programs, and its survey numbers are always the latest 5-year estimates, which is why a city's income on QuickFacts can trail a newer 1-year figure.

For the basics of each program, see our explainers on the purpose of a census, whether the American Community Survey is legitimate, and what a census tract is.

Sources

U.S. Census Bureau, "The American Community Survey and the 2020 Census"; "Comparing ACS Data"; "When to Use 1-year or 5-year Estimates"; Population Estimates Program; "Which Data Tool Should I Use?". Population figures from the 2020 Census, Vintage 2025 population estimates, and 2025 American Community Survey 1-year estimates (tables B02001 and S0201).