Also, this...
Joe Lenski, executive vice president of Edison Media Research and one of the directors of the exit polls, conceded that the surveys, which aimed at questioning every 10th voter leaving 1,480 precincts nationwide, had missed the mark.
"A superficial reading of the data would have indicated that Kerry had a slight lead," he said in a telephone interview. But he defended the survey's final result, calling it "well within the sampling error," which ranged between 2 and 5 percentage points."
"I've been doing exit polling since 1987," he said. "I know the margin of error. The differences we saw in just about every state fell within those calculated errors."
reflector.com
Exit surveys, by their nature, always vary somewhat from the actual vote count. In recent years, they have tended to over-estimate the support for Democratic presidential candidates.
Writing for the Century Foundation, a liberal policy group, Ruy Teixeira tracked past preliminary exit poll numbers and found the Republican vote to be consistently under-estimated before the data were adjusted to reflect actual vote results. In 1988, for example, initial findings showed Democratic candidate Michael Dukakis with 50.3 percent of the vote, although George H.W. Bush won the election.
Another issue is why some of the early results had an over-representation of women, who tended to favor Kerry over Bush.
"There's also an issue of whether Democrats were more willing to talk than Bush voters," Frankovic said. She added that experience shows that older voters are generally less willing to participate in exit polls than younger people.
Other unknown factors are the interviewers, since experience shows some are far more successful at coaxing voters to answer the survey than others.
"Polls are pretty blunt instruments," Frankovic said. "They are not as precise as I think some people would like to think." |
