An experimental comparison of question forms used to reduce vote overreporting

Via telephone, a national probability sample of 1464 respondents randomly received one of three question versions of a voter turnout question in the months of December, January, and February, following the November 1998 Congressional elections in the United States. The long version form contained memory cueing techniques and face-saving response options, the short form, modeled after the 2000 American National Election Studies (ANES) question, included only the face-saving response options, and the standard form, modeled after ANES questions used before 2000, included neither. The long form led to significantly lower reported turnout in comparison to both the short and standard forms, indicating that the long form successfully reduced vote overreporting in comparison to the other question versions.