Conducting the Dicennial Census of Population and Housing every ten years is a massive undertaking. Results inform everything from local redistricting and federal funding to social science research that helps us improve health and wellness in our communities.
The first 2020 Census results are expected to be published in August, but a new strategy for protecting individual privacy has demographers, social scientists and state governments concerned. In April, Alabama sued the U.S. Census Bureau over its implementation of “differential privacy,” and 16 states, including Alaska, have signed on.
The Census Bureau is trying to deliver on dueling congressional mandates, said Dr. Matthew Berman, Professor of Economics at ISER, in a recent presentation. On the one hand, the Bureau must keep individuals’ data confidential for seventy years. On the other, it must deliver accurate population counts. Advances in technology have made it no longer possible to do both, according to Berman.
Maintaining privacy is becoming more difficult in a world of big data. Private companies have been gathering their own information on the populace from the internet – scraped from social networks and government documents such as voter registration, court records, and property ownership databases. Today it’s possible to find a lot of information about individuals with just a quick search.
Differential privacy protects the information of specific individuals by adding random noise, especially to sub-populations where people could be easily identified. The agency provided a demonstration of this approach, showing how it would have affected the 2010 census numbers. Berman took a look at the Alaska data. “The data are not usable” when differential privacy is used, said Berman, who found that Alaska Natives and American Indian populations were undercounted in rural communities.
Overall counts, like state or national population figures, are still fairly accurate. It’s when you look at detailed breakdowns of the population that the effects of the random errors are apparent. “When they take numbers from one community, they have to put them somewhere,” said Berman. On the demonstration site he found that the Census had “moved” people from inhabited villages into abandoned villages and seasonal houses and cabins. “In Alaska we know that no one lives there, but if you’re just looking at Google Earth, you wouldn’t know it’s abandoned.”
This illustrates how differential privacy could affect voting rights, because the Census Bureau couldn't prove that the numbers accurately count the number of minorities in a certain area that is included in a gerrymandered district, legislative district, or even congressional district.
The census tells us who we are and where we are going as a nation, and helps our communities determine where to build everything from schools to supermarkets, and from homes to hospitals. It helps the government decide how to distribute funds and assistance to states and localities. It is also used to draw the lines of legislative districts and reapportion the seats each State holds in Congress.” – U.S Census Bureau.
In April 2021, Alabama was the first state to file a lawsuit over this issue. Several other states have signed on, including Alaska. “We knew it was coming,” said Berman, who predicts this may go to the Supreme Court. He believes it’s really an issue for Congress and the administration. “The litigation is a wake-up call for policy makers that they need to get involved.”
Looking to the existing Federal Policy for the Protection of Human Subjects (“Common Rule”) may provide a solution, Berman noted. Federally funded researchers who work with human subjects must apply strict ethical guidelines, such as obtaining informed consent, considering risk of harm from inadvertent disclosure of confidential information, and using independent Institutional Review Boards to review and approve plans, and oversee implementation of their research.
Differential privacy addresses the trade-off between privacy and accuracy, while the Common Rule addresses the trade-off between the risk of harm from disclosure of private information and the public benefit from the information collected. “The public benefit really has to be addressed by our elected officials,” said Berman.
More information
Data Privacy and the 2020 Census: state lawsuit shows why Congress should act, Matthew Berman, Anchorage Daily News, June 2, 2021
Census Bureau Data: Balancing Privacy and Accuracy, ISER webinar with Dr. Matthew Berman, May 17, 2021
Census Bureau webinars on differential privacy
16 states back Alabama’s challenge to Census privacy tool, AP News, April 13, 2021