In an April 18 congressional hearing, the Census Bureau’s acting director, Ron Jarmin, gave his first public testimony since a citizenship question was added to the 2020 census. During the hearing, Congressmembers raised concerns about decreased response rates among noncitizens, who may fear how the data will be used. Though the Census Bureau cannot share individual-level information with other federal agencies, including Immigration and Customs Enforcement (ICE), Congressmember Grace Meng (D-N.Y.) noted that census block-level data (which are publicly released) were used during World War II to target Japanese-Americans for internment.

Jarmin testified that people who do not fill out the citizenship question will still be counted. During a hearing before a Senate subcommittee on April 25, Attorney General Jeff Sessions also said that the controversial census citizenship question would come last on the form and that “it shouldn’t scare people. They don’t have to answer it, really. . . . I believe the concerns over it are overblown.”

In actuality, refusing to answer a census question can result in a fine or home follow-up visit by Census Bureau workers. If noncitizens respond to the census and answer the citizenship question, they could be targeted for deportation using publicly-available block-level data. If they do not respond at all, areas with large noncitizen populations could be undercounted, giving them less political representation and federal funding.

The administration has justified including the citizenship question by highlighting the benefit of having more data to enforce the Voting Rights Act. However, data scientists and activists have major concerns over the data’s use and its possible lack of accuracy.

Background

The census is a sample of 100 percent of the population, done every ten years, and has not included a citizenship question since 1950. Instead, the question was moved to the “long form” survey, a 17 percent sample conducted every ten years from 1970 to 2000. The “long form” was replaced by the American Community Survey (ACS), a 2.6 percent sample done every year, with a citizenship question since 2005.

In December, 2017, the Department of Justice wrote that the ACS citizenship data were not precise enough to enforce the Voting Rights Act. They requested a citizenship question be added to the census.

Commerce Secretary Wilbur Ross has the authority to determine which questions to include on the census. In a memo released in March, Ross considered concerns that adding a citizenship question could reduce response rates for noncitizens. He asserted there was no evidence the new question would materially reduce responses.

But the citizenship question is understudied

The citizenship question “has never been tested in a contemporary census environment, so of course, there’s not going to be any evidence,” John Thompson, the previous census director who resigned last year, told Science in April. Thompson is one of six former Census Bureau directors who signed a letter to Ross criticizing the decision earlier this year. The American Statistical Association has cautioned that “it is unsound practice to make last-minute, untested changes to such an important activity.”

Why include a citizenship question?

Citizenship data from the census’ 100 percent sample could give precise information for cases involving redistricting (usually done after each census). Other cases require citizen voting-age population data in non-census years, and can rely on the ACS (conducted yearly). ACS estimation could be made more precise by using citizenship data from the census.

Why not include a citizenship question?

There are two concerns: data use and data accuracy.

Congressmember Meng raised the concern that publicly-released census block-level data could be used to target noncitizens. For a sense of scale, the United States has 11,078,297 census blocks, an average of 30 people in each. ICE or vigilantes could use this fine-grained information to find noncitizens.

Fear of being targeted for deportation may cause noncitizens to not respond accurately to the census, either by not participating or by marking “citizen,” which Ross noted happens in the ACS about 30 percent of the time. Inaccurate data on the citizen voting-age population could muddle Voting Rights Act cases that the question was requested to support.

Furthermore, noncitizens not responding to the census could result in an inaccurate count of the total population. This could distort allocation of congressional seats and three-quarters of a trillion dollars in federal funding for public health, education, transportation, and much more. Undercounting noncitizens could lower political representation and federal funding for areas with large noncitizen populations. Surveys often rely on census totals to get better estimates. The ACS estimates use census totals, as do the Small Area Income and Poverty Estimates (SAIPE), the basis for annual Title I allocations of federal funds to school districts. Inaccurate census data could distort these and many other policy-relevant survey estimates.

In his memo, Ross emphasizes precision gains from having citizenship data in the census’ 100 percent sample. He deemphasizes accuracy concerns by pointing to the absence of evidence that the new question would materially reduce response. But this lack of evidence is due to a lack of study. In addition to the statistical concerns of precision and accuracy, there are serious worries about data use. History tells us that census data can be used to harmfully target people. Given these high stakes, advocates for adding questions should have the burden of proof to show that the benefits outweigh the harms.

Those who want a census that ensures that noncitizens and their neighbors have access to political representation, federal funds and protection from deportation have avenues for action: lawsuits like the one filed by the Black Alliance for Just Immigration in California this April against the Commerce Department (which oversees the Census Bureau) and mobilizing as many people as possible to respond to the census and skip the citizenship question on the survey.

This is reader-supported news. Make a contribution today!

____

Photo: A lawn sign promoting the 2010 census. Credit: U.S. Census Bureau.