ERRORS IN MEASURING VOTE CHOICE IN THE NATIONAL ELECTION STUDIES, 1952-88

This study examines a problem of systematic measurement error in the National Election Studies (NES). A significant number of respondents overstate their support for the winning candidates. This bias for the winner in vote reports has been a problem at some time for the four offices examined in this analysis: modest overall for the presidency, serious throughout for senator and governor, and serious in House elections only since 1978, when the NES changed the form of the vote choice question. Variations in the winner bias are documented for conditions of question wording changes, voter validation, and incumbency. The effects of interview timing are also shown to be significant in some cases but to vary by office. The problem of biased vote reports has a serious distorting effect when estimating a large class of vote choice models in subpresidential elections. The study concludes with a brief discussion of some of the options available for dealing with the problem.