Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: The accuracy of the 2004 pre-election and exit polls are examined from a total survey error perspective. There was considerable speculation before the election that Republicans were being oversampled, partly because of the exclusion of people who only have cell phones from sampling frames. Weighting by partisanship seems to have accidentally overcompensated for that problem. By contrast, Republican voters were less likely to be interviewed in exit polls, partly because of interviewer-related effects. All in all, the problems with the exit polls were more serious than those for the pre-election polls in 2004.