A new study from Georgetown University indicates that the addition of a citizenship question to the 2020 Census is "likely to increase the costs and compromise the accuracy of the 2020 Census by increasing nonresponse."

The question asked on the American Community Survey (ACS) is a "quite different" context than the decennial Census, "but analysis of the ACS citizenship question can shed light on potential responses to the citizenship question in the 2020 Census," according to the new study from William O'Hare at Georgetown Law School’s Center on Poverty and Inequality.

An analysis of the nonresponse rates for the citizenship question on the ACS shows that it "is much higher than the nonresponserates for any of the other questions that will be on the 2020 Census questionnaire." In 2016, it was 6 percent, and no other question planned for the decennial questionnaire had a higher rate than 1.8 percent. Worryingly, "nonresponse rates for the citizenship question have increased since 2010 while the nonresponse rates for other questions that will be on the 2020 Census have remained stable."

Nonresponse rates to the citizenship question vary dramatically across the country, with the 9 percent nonresponse in Arizona versus 2.6 percent in Vermont. Eleven cities have only 3 percent or less nonresponse, but 43 cities hit 10 percent or more. Unsurprisingly, variance by demographic group means that the highest nonresponse rates are found among racial and ethnic minorities, the foreign born, and residents of central cities of metropolitan areas. Finally, data collection mode matters: nonresponse rates are higher via "self-response" than via in-person nonresponse follow-up interviews (8 percent for Internet, 6.7 percent for mail, and 3.8 percent for in-person).

The Department of Commerce decided to add to a citizenship question the 2020 Census questionnaire earlier this year, a move opposed by the Insights Association and more than 250,000 other groups (according to estimates from the Leadership Conference).