|
Abstract:
|
In 2010, automated matching of name and date of birth for every census enumeration will better identify duplicates than in past censuses, but the problem of omissions persists. Ironically, eliminating duplicates might result in a larger net census undercount. This research evaluates new questions to reduce omissions. At the end of the short form, respondents are given reminders of people who might be missed, and asked to review the form and answer 2 questions about possible errors. The questions provide a clear stopping point in the census form, which helps avoid confusion. In a March 2006 national test, experimental and control forms are mailed to random samples of 7,100 households each. Returns are analyzed to determine whether (1) respondents answer the questions, (2) a final review reduces errors, and (3) the questions flag errors, as determined by a follow up interview.
|
