PRB

Data products from the 2020 Census differed from previous censuses. Many of the changes were intended to improve the privacy of individual responses, but those improvements come with challenges.

We’ve heard from many data users asking about the effects of the new disclosure avoidance procedures on 2020 Census data. Over the past three years, PRB and U.S. Census Bureau staff have written a handbook and a series of briefs to help make sense of privacy protections in the decennial census. Here are the top 10 things to know:

1. The Census Bureau applied new privacy protection techniques to 2020 Census data to protect respondent information against increasingly sophisticated attacks.
2. The new privacy protections are based on a mathematical framework called “differential privacy.” It works by adding statistical noise—small, random additions or subtractions—that make it difficult to link the published data with other sources to identify individuals.
3. Data for very small counts, such as population in a census block, may be noisy. In most cases data users should aggregate data into larger geographic areas before use.
4. The noise added to data published from the 2020 Census is comparable to or smaller than the magnitude of other sources of error in the data, such as operational and coverage error.
5. Although all 2020 Census data products use the differential privacy framework, the way the framework was applied varies. For more information on these applications and their strengths and limitations, refer to the briefs listed below.
6. 2020 Census data often do not match across data products, and sometimes do not match across different tables within data products. Use the data table most closely related to your topic of interest and refer to the briefs below for more information.
7. Some illogical values, like negative counts, were removed from the data—either through adjustment or suppression. However, other illogical results remain in the data—such as blocks with resident children under the age of 18, but no adults. Aggregating data—either by geography or by population group—is often helpful at eliminating the illogical results, but data users should consult the briefs for specific guidance.
8. The Census Bureau processed population and housing data separately in all data products (except S-DHC). Separate processing means that combining person and household data may lead to illogical results, such as more occupied households than people in a geographic area. For most applications, aggregating data to larger geographic or demographic categories resolves this issue.
9. Some 2020 Census counts were reported as enumerated, with no noise added. These include the following:
   1. Total number of people in each state, the District of Columbia, and Puerto Rico.
   2. Total number of housing units (but not population counts) in each census block, and at all other geographic levels.
   3. Number of occupied group quarters facilities (but not population counts) for selected facility types.
10. If you have questions about disclosure avoidance in the 2020 Census, contact the Census Bureau 2020DAS@census.gov.

The Census Bureau is using lessons learned from the 2020 Census to help inform the plan for data products in the 2030 Census. Subscribe to the Census Bureau’s newsletter to stay to up to date on plans for next census.

For more information about disclosure avoidance in the 2020 Census, refer to the following resources:

• Why the Census Bureau Chose Differential Privacy
• Disclosure Avoidance for the 2020 Census: An Introduction
• Disclosure Avoidance and the 2020 Census Redistricting Data: What Data Users Need to Know
• Disclosure Avoidance for the Demographic and Housing Characteristics File (DHC) and Guidance for Data Users
• Disclosure Avoidance and the 2020 Census: How the TopDown Algorithm Works
• Disclosure Avoidance and the 2020 Census: How SafeTab-P Works
• Disclosure Avoidance and the 2020 Census: How SafeTab-H Works
• Disclosure Avoidance and the Supplemental Demographic and Housing Characteristics File (S-DHC): How PHSafe Works