U.S. Census Bureau Releases New 2021 American Community Survey (ACS) 1-Year PUMS and ACS 1-Year Supplemental Estimates

2021 ACS 1-Year Public Use Microdata Sample (PUMS) Files 

We are pleased to announce the release of the 2021 American Community Survey (ACS) 1-year Public Use Microdata Sample (PUMS) files. The PUMS files show the full range of population and housing unit responses collected on individual ACS questionnaires for a subsample of ACS housing units and group quarters persons (approximately one percent of the United States population), with disclosure protection enabled so that individuals or housing units cannot be identified. 

The PUMS files allow data users to conduct a custom analysis. Working with PUMS data generally involves downloading large datasets onto a local computer and analyzing the data using statistical software such as R, SPSS, Stata, or SAS. 

PUMS data are currently accessible via the ACS website, and the FTP site, and the microdata analysis tool on data.census.gov. Data.census.gov is particularly useful for researchers who need quick statistics with PUMS.  

For more information about the PUMS files, visit the links below:

2021 ACS 1-Year Supplemental Estimates 

We are also pleased to announce the release of the 2021 ACS 1-year Supplemental Estimates.

Supplemental Estimates are simplified versions of popular ACS tables for geographic areas with at least 20,000 people, compared to the 65,000 population minimum for the standard ACS 1-year estimates. 

Visit the Census Bureau's data.census.gov or use the Census API to start exploring these estimates.

For more information on ACS Supplemental Estimates, visit the links below:  

Parents
  • OK, I've read the chart ACS1 vs update vs ACS5 and watched a few videos. Can I assume (yes never a good idea)? That if I'm looking at a field \ measure median rent for a 2 bedroom for example. I notice in the ACS1 that there are counties where the data is not there 999999999.  They do block data where the sample size is too small (may disclose personal data etc.).  If I look at the same geo area (county level for example) which was a 999999 on the ACS1 it would also be a 9999999 on the supplemental and also on the ACS5?  Has anyone done any analysis around this issue or question? Thanks!

  • Since suppression is in part based on numerical thresholds ("minimum number of cases" in Bureau parlance), I would expect that counties suppressed in the ACS 1-year would also be suppressed in the 5-year and the supplemental. But I haven't done any analysis to prove it.

    See https://www2.census.gov/programs-surveys/acs/tech_docs/data_suppression/ACSO_Data_Suppression.pdf for the details of suppression, particularly section 2.5.1, Minimum Number of Cases.

  • I did not know the fields and names would be different ACS1 vs SCA supp, making a comparison difficult if not impossible. I took S2701_C01_017E Estimate!!Total!!Civilian noninstitutionalized population!!RACE AND HISPANIC OR LATINO ORIGIN!!Black or African American alone which has 295 nulls among the 800 counties in ACS1. Then took K200201_003E Estimate!!Total:!!Black or African American alone. from ASC Supp.

    Most nulls (90%)  in the ACS1 had data in ACS supp. However, when they both had data, the data values usually were within 5-10% of each other, but some were 30% diff from the ACS1 and ASC Supp.  This is way off. So I can pull apples to apples (maybe my comparison wasn't perfect?) and get a 30% difference...

    If anyone here could check; Black or African American alone  (ACS1 vs ACS Supp) any way you feel is best, and let me know if you see any differences in these counties. It would help me to know if I'm doing something fundamentally wrong.

    LaPorte County, Indiana
    Coryell County, Texas
    Monterey County, California

  • Ok, I did more digging and in the 3 examples I gave, all the data / fields are significantly different from the ACS1. So the actual variance between race or age as a percent of the total is off by under 5%.  As the ACS5 will be out soon, and it has the same structure of ACS1. I think the way to go is ACS1 or ACS5. Also, Microdata seems really great except it doesn't do counties, which is my focus right now.

Reply
  • Ok, I did more digging and in the 3 examples I gave, all the data / fields are significantly different from the ACS1. So the actual variance between race or age as a percent of the total is off by under 5%.  As the ACS5 will be out soon, and it has the same structure of ACS1. I think the way to go is ACS1 or ACS5. Also, Microdata seems really great except it doesn't do counties, which is my focus right now.

Children
No Data