Skip Header

ACS Updates on Disclosure Avoidance and Release Plans

August 20, 2020
Written By: Dr. John M. Abowd, chief scientist and associate director for Research and Methodology, and Donna M. Daily, chief, American Community Survey Office

Despite changes and delays during this unprecedented time, we are happy to report that the U.S. Census Bureau is on track to release the 2019 American Community Survey (ACS) 1-year estimates as scheduled Sept. 17, 2020. Check out our website for the full data release schedule and other details.

As we prepare to release the 2019 ACS products, we want to remind data users of our commitment to protect respondent privacy and confidentiality. Prior blogs outlined steps the Census Bureau is taking to modernize the procedures we use to protect respondent data. Our adoption of formal privacy will allow us to strengthen safeguards and increase transparency about the impact of privacy protections on data accuracy.

As Deputy Director Ron Jarmin previously stated, we do not plan to implement formal privacy for the full suite of ACS data products before 2025. In the interim, we will continue to evaluate existing privacy protections and bolster them as necessary to address privacy risks that emerge. Our goal is to maintain the utility of the ACS as the preeminent federal survey for federal, state and local data users, while remaining committed to our legal obligations to protect confidentiality.

One area where we are strengthening our disclosure avoidance methods is the count of final interviews published in our quality measures and other detailed tables. We will continue to publish the quality measures tables that provide this information for select geographies. But we are discontinuing the tables that included this information for all geographies.[1]  In addition, we are adding “noise” to the interview counts to cut the risk of disclosure while still providing a general indicator of data quality for the geography of interest. We will continue to publish the household sample sizes selected for invitation to complete the ACS, without added noise or rounding. The reason: These sample sizes are properties of the ACS design, not its realized sample, and provide a more robust indicator of data quality for very small geographies.

Table B98001 (Unweighted Housing Unit Sample) will be published for nation, state, county and place in both the 1-year and 5-year ACS beginning with the 2019 ACS release. Previously, the table was not published for places in the 1-year ACS. A new table will be created of final person interviews, which will be released for nation, state, county and place in the 1-year and 5-year ACS. This will provide the same information as the former B00001, but will be restricted to the same summary levels as B98001.

The Census Bureau has a tradition and public expectation of producing high quality statistics while protecting the confidentiality of respondents. We will work closely with our scientific and data user communities as we explore options for modernizing ACS privacy protections while ensuring the data products’ continued high quality and fitness-for-use. The Census Bureau is funding collaboration opportunities with external researchers on the issue of formal privacy for sample surveys. A key deliverable of such collaboration will be establishing effective data user engagement. We will provide more information about this effort as it becomes available.

Send comments or questions to <ACSprivacy@census.gov>.

 

[1] These data were published in B00001 and B00002 in the 1-year and 5-year ACS products as well as K200001 and K200002 in the 1-year supplemental ACS product.

 

 

Back to Header