Boston Area Colleges Election Project

Preelection polls for the 1989 VA gov. election overestimated the vote share of the black candidate. This was due to social desirabilty and race-of-interviewer effects

To be accurate, exit polls need to be conducted throughout election day. Different groups vote at different times of day.

African American respondents performed more poorly in answering factual questions (rather than questions on opinion) when interviewed by a white interviewer than when interviewed by an African America

Giving voters secret ballots in exit-polling leads to higher response rates and more accurate results than doing face-to-face interviews

Overreporting in the NES has gotten progressively worse as a result of the sample being over saturated with self-reported voters

Race of interviewer has an impact on responses to racial questions when the respondent is white and the interviewer is black.

Black candidates David Dinkins and Douglas Wilder did better in the polls than in the actual returns in the 1989 elections, similar to Tom Bradley's experience in the California race for governor in 1

Blacks interviewed by whites were more likely to express closeness toward whites than blacks interviewed by blacks.

The Loyola Marymount poll, the poll implemented in this study, used a "racially stratified homogonous precinct approach that specifically designated precincts in predominantly White, Black, Latino, a

Examines failures of pre-election polls to predict accurately the Clinton/Obama primary results. Submits that ultimately, likely-voter models are the culprit of the inaccuracies.