Akshay DineshA town can gain renters, lose manufacturing jobs, add longer commutes, and see household incomes rise between two national censuses. Waiting ten years to notice those changes would leave school districts, transportation planners, businesses, and residents working with an old picture. The American Community Survey, or ACS, fills that gap by producing fresh estimates about people and housing every year.
The U.S. Census Bureau released the 2025 ACS one-year estimates on October 29, 2026, covering topics from language and education to rent, income, health insurance, and commuting. Those tables are extraordinarily useful, but they are not a simple annual head count. They come from a sample, cover a period of time, and include uncertainty. Reading them well means knowing what kind of estimate is on the screen and what claims it can actually support.
What the American Community Survey measures
The decennial census aims to count every person in the United States at a specific point in time, April 1 of a census year. The ACS has a different job. It describes the characteristics of communities: how people travel to work, what they pay for housing, which languages they speak at home, what education they have completed, whether they have health insurance, and dozens of other subjects.
Rather than contact every household, the Census Bureau selects a scientific sample of addresses throughout the year. Roughly 3.5 million addresses are chosen annually, divided into monthly samples. A selected household may respond online or by mail; some addresses that do not respond are chosen for follow-up interviewing. The survey also includes people living in group quarters such as college residence halls, nursing facilities, and correctional institutions.
Each completed response represents more than one household or person. The Census Bureau assigns statistical weights, adjusts for missing responses, and aligns the estimates with independent population and housing controls. That process turns a sample into estimates for the country, states, counties, cities, congressional districts, and other geographic areas. It also explains why an ACS population estimate may not match a decennial census count or a separate population estimate.
One-year and five-year estimates answer different needs
ACS data are published in one-year and five-year products. The names describe the period of data collection, not how long the numbers remain valid. A 2025 one-year estimate combines responses collected from January through December 2025. A 2021–2025 five-year estimate combines 60 months of responses across those five calendar years.
One-year estimates are the more current option, but the smaller sample gives them more statistical uncertainty. The standard one-year product is available only for areas with populations of 65,000 or more. A simplified set of one-year Supplemental Estimates reaches areas with at least 20,000 people. Five-year estimates cover every geography, including census tracts and block groups, because pooling five years of responses creates a larger sample.
The right choice depends on the question. To compare recent conditions among large states or metropolitan areas, one-year estimates may be appropriate. To study a small town, a neighborhood, or a narrow population group, five-year estimates are usually necessary and often more reliable. The tradeoff is currency: a five-year estimate describes the entire period, not just its final year.
That last point prevents a common mistake. A 2021–2025 estimate is not a measurement taken in 2025, and it is not simply the value from the middle year. It reflects answers gathered throughout all five years. It should be compared with another five-year estimate, not with a one-year figure that uses a different sample and time frame.
The margin of error belongs beside the estimate
Every ACS estimate is accompanied by a margin of error because the number comes from a sample rather than the full population. The Census Bureau publishes these margins at a 90 percent confidence level. If a table reports a median gross rent of $1,500 with a margin of error of $80, the corresponding interval runs from $1,420 to $1,580.
The interval does not mean that every rent falls within that range. It describes uncertainty around the estimated median. A large margin of error compared with the estimate signals that the result is less precise, which often happens for small places or narrowly defined groups. Percentages, medians, and totals all have their own margins of error.
Suppose two counties have estimated college-graduation rates of 31 percent and 33 percent. It is tempting to declare that the second county has the higher rate. If both estimates have sizeable margins of error, however, the observed difference may be consistent with ordinary sampling variation. The Census Bureau provides a statistical testing tool for comparing ACS estimates and their margins of error; looking only at which point estimate is larger can overstate the evidence.
Uncertainty does not make the data useless. It tells readers how cautiously to phrase a finding. “The estimate is higher” is a description of the table. “The community truly has a higher rate” is a stronger claim that needs a statistically meaningful difference and a comparison built from compatible data.
Four comparisons that can mislead
Most ACS mistakes begin with a reasonable-looking comparison. The figures may sit in the same table or share the same topic, yet their underlying periods, populations, or definitions differ. Four checks catch many of the problems.
- Do not mix one-year and five-year products. Compare one-year estimates with one-year estimates and five-year estimates with five-year estimates. Their sample sizes, precision, and periods are different.
- Be careful with overlapping five-year periods. The 2020–2024 and 2021–2025 estimates share four years of data. A small difference between them is not the same as a clean year-to-year change. Nonoverlapping periods, such as 2016–2020 and 2021–2025, provide a clearer comparison when definitions and boundaries remain compatible.
- Check the universe and denominator. A table about workers age 16 and older does not describe every resident. A percentage of occupied housing units is different from a percentage of all housing units. The table title, universe line, and notes define who or what is included.
- Check changes in questions and geography. Survey wording, table definitions, and geographic boundaries can change. A county subdivision, school district, or congressional district may not represent the same area in two releases. The Census Bureau publishes user notes, geography changes, and table modifications for each data year.
Rates and totals also tell different stories. A county may add residents living in poverty even while its poverty rate falls if the total population grows faster. Likewise, comparing dollar amounts across many years may require an inflation adjustment. The ACS can reveal change, but the comparison must match the claim.
A dependable way to use data.census.gov
Begin with a specific question rather than opening the largest table available. “What share of households in this county lack a vehicle?” is easier to answer correctly than “What does the data say about transportation?” Define the geography, population, measure, and time period before choosing a product.
- Choose the correct dataset. Use a one-year product for current comparisons among sufficiently large areas and a five-year product when smaller geographies or greater precision matter.
- Start with a profile when exploring. Data Profiles and Subject Tables summarize major measures. Detailed Tables offer more categories and are better once the exact concept is clear.
- Read the table title and universe. Confirm that the rows describe the intended population, households, workers, or housing units.
- Keep the margin of error visible. Download both estimates and margins of error, especially when comparing places, years, or subgroups.
- Record the table ID, dataset, year, and geography. A statement is easier to verify when its source is specific, such as a 2025 ACS one-year table for a named county rather than simply “Census data.”
- Check notes before calling something a trend. Look for changes in wording, methodology, boundaries, inflation treatment, and overlapping periods.
The ACS is powerful precisely because it gives communities information that a once-a-decade count cannot provide. Its estimates can show where rents are rising, commutes are changing, languages are being spoken, or insurance coverage is shifting. The numbers become trustworthy evidence when their time period, population, and uncertainty travel with them. A careful reader does not ask only, “What is the estimate?” but also, “Which survey product produced it, whom does it describe, and how certain is it?”
