Skip to content Skip to main navigation Report an accessibility issue

Differential Privacy and the 2020 Census

Nicholas Nagle, Ph.D.
Associate Professor of Geography
University of Tennessee, Knoxville

Originally presented May 21, 2020.

2020 Census results are confidential, but what if today’s computing power could be harnessed to re-identify individuals who responded? That’s the challenge facing the Census Bureau and the proposed solution called differential privacy could potentially impact the accuracy of statistics it will report next March.

Differential privacy might seem like a wonky topic but we promise it’s very important. From what we know thus far, it could result in significant changes to the 2020 Census data products. The wide-ranging discussions include reducing the amount of data and the number of tables. Some questions include:

  • What if noise injected into 2020 Census data rendered certain geographies-like census blocks-statistically unreliable?
  • What if data about race, ethnicity, household size and other statistics were available only at higher levels of geography, such as block groups or tracts?
  • What if the population of cities and counties had a margin of error? Does this create a lottery for each city’s distribution of state-share tax revenues in Tennessee?
  • How can communities conduct post-census challenges-like Census Count Question Resolution and Special Census programs-when error is intentionally infused in the data?

We are fortunate to have Dr. Nagle from the Geography Department at the University of Tennessee, Knoxville, addressing these questions and more in the webinar next week. He sits on a National Academy of Sciences panel that advises the U.S. Census Bureau about differential privacy, and chairs a census geography subcommittee about the topic as well.

Dr. Nagle’s article published in the Conversation this spring is a good introduction to the subject and highlights one particular area of concern: state-shared revenue distribution to counties and municipalities.