Wednesday, November 6, 2019

Test run of Differential Privacy for the 2020 Census

The Census Bureau is investigating a technique, known as Differential Privacy, to avoid inadvertently disclosing individual information in the 2020 Census. (See the Census Bureau's overview for more on how differential privacy works.)


But some data users are concerned that the level of noise added to the data, or the process used to add that noise, may result in poor data quality.

To give data users a chance to assess how Differential Privacy (DP) would affect the quality of data, the Census Bureau released demonstration data based on the 2010 Census. The demonstration data can be compared with original 2010 Census tables to see how much "noise" DP would introduce. Files are available via FTP download and require SAS or other statistical programming to open the files and extract the data.

For users who do not have access to the statistical software or coding skills required to work with the raw demonstration data tables, IPUMS NHGIS produced user friendly data tables that can be accessed on their website.

The Committee on National Statistics has also issued a Call for Input for their December Workshop on 2020 Census Data Products: Data Needs and Privacy Considerations.

No comments:

Post a Comment

your insights?