The survey experiment from researchers at Harvard Kennedy School’s Shorenstein Center on Media, Politics and Public Policy found that asking the question could result in around 8.4 % of the Hispanic population recorded in the 2010 census being missed.

Qualtrics recruited the survey panel of around 9,000 people and conducted the study in two waves: the first targeted at non-Hispanic racial groups and the second targeted at Hispanics.

The study was designed to mirror the form and content of the US census, based on its proposed questions. It found that asking respondents about citizenship significantly reduced the overall number of questions answered by all respondents and significantly reduced the number of household members reported as being Hispanic.

The researchers said they have likely underestimated the effect of the question. "Not only are we university affiliated academic researchers – and not the US Government – and so respondent concerns over providing the government with personal information may not have inhibited participation in this survey, but our respondents were paid panelists and thus financially incentivised to complete the survey." 

The Trump administration’s proposal to include a question on citizenship for the first time since 1950 has raised concerns that the question would result in an undercount in rural areas and areas with large immigrant populations.

Federal judges in California and New York have already blocked the question, and the US Supreme Court is set to hear an appeal on 23rd April.