The American Community Survey (ACS)

Introduction to the ACS and basic information about the data.

The American Community Survey (ACS) is a nationwide survey that collects data on the social and economic characteristics of the U.S. population—education, housing, jobs, and more. ACS data is published every year, and survey results are pooled throughout the whole year. ACS estimates reflect data that have been collected over a period of time rather than for a single point in time like the decennial census. Learn more about the ACS.

The American Community Survey is just one of many data programs managed by the US Census Bureau. Learn more about other Census Bureau datasets, and how to ensure you are using the right one.

ACS 1- or 5-year estimates

The ACS publishes 1-year and 5-year estimates each year. You will need to decide which to use. The 1-year estimates use 12 months of survey data to calculate summary tables, while the 5-year estimates use data collected over 60 months. For reference, the 2019 1-year ACS total population sample for San Francisco County included just over 11,000 individuals and the 2020 5-year sample about 54,000 (between 1% to 6% of the total population). The 5-year estimates include the data from the four prior years to achieve that sample size.

If you are showing a data table for all of San Francisco (with no further breakdowns or sub-sections), the 1-year estimates are likely preferable for timely and relevant data. If you are doing any cross sections or subdivisions of San Francisco, you will likely need to use the 5-year estimate to be statistically reliable. Examples of cross sections would be demographic data for census tracts, or looking at the intersections between 2+ variables – like homeownership rates by race in San Francisco. This is because with each slice of the data, the survey sample becomes smaller and the reliability of the data decreases. Learn more about the tradeoffs between 1- and 5- year estimates.

ACS year comparisons

When comparing two estimates from the ACS, or tracking trends over time, compare non-overlapping datasets (example: compare 2005-2009 ACS 5-year estimates to 2010-2014 ACS 5-year estimates). Do not compare estimates from overlapping datasets (example: do not compare 2005-2009 ACS 5-year estimates to 2006-2010 ACS 5-year estimates). This is because the two samples will not be independent. The same survey respondents will be represented in both overlapping datasets so you cannot make a statistically accurate comparison. Read more about what years to compare from the ACS.