Harvard researchers come out against Census privacy technique

Harvard University researchers are voicing opposition to a privacy tool used by the U.S. Census Bureau that obscures information used to draw out congressional districts, arguing that it makes it extremely difficult to make sure districts have equal populations. 

In a paper released last week and reported by The Associated Press on Wednesday, the researchers argued that the privacy method adds “noise,” or intentional errors, to the Census data to obscure the identity of 2020 respondents while still providing information on them. 

The researchers went on to say that the tool makes it “impossible” for states to comply with the equal population requirement, also known as “One Person, One Vote.” 

“Over the past half century, the Supreme Court has firmly established the principle of One Person, One Vote, requiring states to minimize the population difference across districts based on the Census data,” they noted. 

The researchers argued that the tool, called “differential privacy,” in effect “negatively impacts the redistricting process and voting rights of minority groups without providing clear benefits.”

The authors argued that the Census Bureau should instead rely on the privacy methods used in the 2010 census, in which data of some households was switched with others. 

The paper from the researchers, who included political scientists, statisticians and a data scientist, featured a simulated political map with the 2010 Census data using the differential privacy tool. 

The researchers found that the privacy method undercounted racially mixed and politically mixed areas, while also overestimating racially and politically segregated districts. 

Thus, the authors said that the technique “tends to introduce more error for minority groups than for White voters, and even more error for voters who are in a minority group” at the local level. 

The report comes as the Census Bureau is making a final decision on how it will use differential privacy. 

In April, two Census Bureau officials, John Abowd and Victoria Velkoff, wrote in a blog post that privacy protections are needed for census data to prevent hackers from gathering identifying information about participants. 

“With today’s powerful computers and cloud-ready software, bad actors can easily find and download data from multiple databases,” they wrote at the time. “They can use sophisticated computer programs to match information between those databases and identify the people behind the statistics we publish. And they can do it at lightning speed.” 

Federal judges in Alabama are currently considering a legal challenge to the “differential privacy” tool, with state’s attorneys arguing that it results in inaccurate redistricting numbers.