2020 Census results are confidential—but what if today’s computing power could be harnessed to reconstruct responses and re-identify individuals who responded? That’s the challenge facing the U.S. Census Bureau, and the proposed solution—differential privacy—could potentially impact the accuracy of statistics released following the conclusion of the 2020 Census.
For decades, federal laws have been in place to ensure that the statistical data released by the Census Bureau protects the confidentiality of survey respondents. In past decennial censuses, these procedures included swapping households of similar sizes between areas. For 2020, the new disclosure avoidance proposals formalize the privacy protection process through the mathematical injection of noise into response tabulations.
A number of concerns about the new methods have been raised. These include the trade-off between accuracy of the published 2020 statistics, detailed data being provided for fewer Census geographies and potentially fewer data tables.
Introduction to Differential Privacy
Protecting Privacy with MATH
An engaging overview that introduces the challenge of protecting privacy in a world where massive computing power can overcome the Census Bureau’s traditional privacy protections.
Census 2020 will protect your privacy more than ever – but at the price of accuracy
The Conversation by Nicholas N. Nagle, Associate Professor of Geography, University of Tennessee
This article summarizes both the concerns about keeping census data confidential and the impact of supplying differentially private 2020 Census Data.
Differential Privacy for Census Data Explained
National Conference of State Legislatures
Overview of differential privacy and the 2020 Census, as well as letters to the Census Bureau outlining concerns over the impact of the new process.
US Census Bureau Resources
2020 Census Data Products Newsletter
Receive regular updates about 2020 Census Products and disclosure avoidance from the Census Bureau team working on the products.
The Census Bureau is reporting 2020 Disclosure Avoidance System (DAS) developments on this web page.
Resources for Tennessee Data Users
Webinar Hosted by the Tennessee State Data Center
University of Tennessee Associate Professor Nicholas Nagle provides an overview of process, and uses data released in the fall of 2019 to explains how state-shared revenue distribution were affected.
Committee on National Statistics, National Academy of Sciences
View presentations from a variety of disciplines explaining the potential impact of differential privacy. The agenda covered an array of topics include redistricting, planning, demography and health.
In October 2019, the Census Bureau released a set of demonstration products that illustrated an early version of the 2020 Disclosure Avoidance System (DAS). Data quality discussions following the release have led to additional work to reduce the error resulting from the noise infusion.