A growing number of Americans are not willing to disclose their citizenship status on a government survey, according to new research. The finding adds fuel to an already fierce political debate over adding a citizenship question to the 2020 census.
State officials and civil rights groups have sued Commerce Secretary Wilbur Ross, challenging his decision earlier this year to add such a question to the decennial census. Ross’s opponents worry some groups, notably foreign-born residents, will shy away from answering the question because of the current hyperpartisan battle over U.S. immigration policy. That could undermine the accuracy of the constitutionally mandated exercise used to apportion seats in the U.S. House of Representatives and allocate $800 billion in federal funds, they say. The new data appear to bolster that argument by documenting rising nonresponse rates to the question on a related Census Bureau survey.
Citizenship is one of 72 questions on the American Community Survey (ACS), an annual sampling of 3.5 million households that in 2005 replaced the long form of the decennial census. The new study finds that the portion of respondents who did not answer the ACS citizenship question more than doubled between 2010 through 2016, from 2.7% to 6%. In contrast, the nonresponse rates for other demographic questions on the ACS—including race, sex, age, and Hispanic origin—remained constant, at less than 2%.
“This suggests an increased sensitivity to being asked about citizenship,” says Indivar Dutta-Gupta of the Center on Poverty and Inequality at Georgetown University in Washington, D.C., which released the report late last week. “The findings lend support to the conclusions of many experts, including former census directors, state officials, and the National Academy of Sciences, that a citizenship question will increase the risks for the 2020 census,” says the author, demographer William O’Hare, a veteran census data cruncher and consultant based in Baltimore, Maryland.
O’Hare found that nonresponse rates on the citizenship question also varied greatly by geography in ways that could jeopardize an accurate reapportionment. The highest rates—Arizona led at 9% and California, New York, and Colorado all exceeded 7%—are home to large numbers of immigrants. The lowest rates, below 4%, were found in Vermont, West Virginia, and Maine—states with relatively small immigrant populations.
There was also variability by racial and ethnic group. Asian-Americans and Hispanics had nonresponse rates of 8.1% and 7.4%, respectively, whereas the rate for non-Hispanic whites was 5.6%. Some 8.3% of foreign-born residents ignored the question, compared with only 5.7% of those born in the United States.
The mode of response also matters. O’Hare found that nonresponse rates were highest among those who answered the ACS online, at 8%. In contrast, 6.7% of those who mailed back a paper ACS bypassed the citizenship question, and the nonresponse rate was only 3.8% for those filling out the ACS through a personal interview.
That disparity could be a double whammy for the 2020 census, O’Hare says. The internet will be an option for the first time, and Census Bureau officials hope more than 60% of U.S. residents will answer electronically. In addition, he says, the 500,000 fieldworkers hired for short-term duty on the decennial census are likely to be less capable of cajoling reluctant residents to answer any questions they have skipped than the smaller and better-trained workforce deployed for the ACS. In effect, says O’Hare, the Census Bureau “is pushing a mode of data collection and a methodology that results in higher nonresponse rates.”
In a March memo justifying his decision, Ross asserts that his staff found “limited empirical evidence” to support the argument that “adding a citizenship question would decrease response rates materially.” But a memo from John Abowd, the Census Bureau’s chief scientist, casts doubt on Ross’s assertion by detailing the negative impact of such a question on response rates. The memo was disclosed as part of the department’s response to the various lawsuits, and O’Hare’s analysis reinforces its message that the citizenship question represents an added burden for some respondents.
“It’s not a random sample,” Dutta-Gupta says about who is more likely to ignore the citizenship question. “The differences [in nonresponse rates] are concentrated geographically and racially. And that’s important.”