How the Census ACS Works: What the Data Actually Measures

Every number on this site comes from the American Community Survey, the largest ongoing demographic survey in the world. Understanding what it is, and what it isn't, is the difference between using census data correctly and drawing bad conclusions.

The Big Shift, from Long-Form Census to Rolling Survey

Until 2000, the US Census Bureau ran two surveys every decade: the short-form decennial census that counts everyone, and a long-form survey sent to one in six households asking detailed demographic, economic, and housing questions. In 2010 the long form was replaced by the American Community Survey, or ACS, a continuous survey sent to a rolling sample of addresses every month.

The change mattered. Instead of detailed data updated once every ten years, the country now gets annual updates with a rolling 5-year smoothing window. The tradeoff is a slight decrease in sample size at any single point in time, in exchange for freshness and consistency.

How the Sample Works

Every month, the Census Bureau mails an ACS questionnaire to roughly 295,000 addresses. Households that don't respond online or by mail get a phone follow-up, and a portion receive an in-person visit. Response is mandatory by federal law, though enforcement is rare. Participation rates are typically in the mid-90% range by the end of the follow-up process.

Over five years, roughly 17.5 million housing unit addresses are sampled. That's a larger sample than any other survey in the world. The massive sample size is what lets ACS 5-Year estimates support reliable data for small areas: cities under 20,000 people, rural counties, individual ZIP codes.

1-Year vs 5-Year Estimates

1-Year estimates are published for geographies of at least 65,000 residents. They're the freshest possible ACS data but cover only larger areas. For small cities and most counties, they don't exist.

5-Year estimates combine sixty consecutive monthly samples into a single, statistically reliable dataset. They are published for every Census geography, every state, county, city, ZIP code, tract, and block group. They are what power most programmatic sites, including this one, because they cover the universe of US places.

The tradeoff is a slight lag. A 5-Year estimate labeled 2023 reflects data collected from 2020 through 2023. That's the normal structure and it doesn't mean the data is out of date, it means it's smoothed across five recent years.

What the ACS Measures

• Population and age. Total counts, age distribution, median age, male and female breakdowns.
• Race and Hispanic origin. Self-reported across seven race categories plus Hispanic/Latino origin as a separate question.
• Income and poverty. Median and mean household income, per-capita income, poverty rate.
• Housing. Median home value, median rent, owner-occupancy, vacancy.
• Education. Bachelor's degree rate, high school graduation rate.
• Commute. Drive-alone, carpool, public transit, work from home, mean travel time.
• Industry and occupation. Employment by industry sector.
• Language, citizenship, disability, veteran status. Social and community characteristics.

What the ACS Does Not Measure

• Births and deaths. Vital statistics come from the CDC National Vital Statistics System, not the ACS.
• Income above a top code. ACS income is truncated to protect privacy. Aggregate statistics still reflect full income.
• Religion. The Census Bureau is prohibited by federal law from asking about religious affiliation.
• Political affiliation. Not collected.
• Small-area change over time. The 5-Year smoothing window means year-to-year change in small areas is often noise, not signal.

Margin of Error

Every ACS estimate has a margin of error, reported alongside the point estimate in the official Census Bureau tables. For large places, the margin is small. For small places, it can be meaningful. When comparing two nearly identical values, look at the MOE before concluding there's a real difference.

How This Site Uses ACS Data

Popreview uses ACS 2023 5-Year estimates across all state, county, city, and ZIP code pages. Data is sourced directly from Census Bureau API tables, not third-party aggregators. Historical population context uses Census Bureau decennial counts and intercensal estimates. Vital statistics (birth data) come from the CDC. Read our methodology for details on how each data type is processed.

Where to Dig Deeper

The Census Bureau's data.census.gov is the authoritative source for every ACS variable. The Bureau also publishes extensive ACS methodology documentation. For aggregated county- and state-level profiles, the IPUMS project at the University of Minnesota publishes cleaned microdata.