Evidence from a randomized control trial indicates that when you introduce a question about citizenship on a United States Census form, respondents answer a smaller percentage of the questions.