According to Febble:
"If the exit poll discrepancy was due to response bias,
the direction of bias was relative undersampling of Bush
voters. About 46% of all voters approached, both Kerry and
Bush voters, refused to be interviewed.
Of these, E-M calculate that to account for the discrepancy
between the exit poll and the vote count, about 47% of the
refusers would have been Kerry voters and about 53% would have
been Bush refusers. Not a great difference, but enough to
swing the poll. You may not accept this hypothesis, but this
is what the "rBr" hypothesis is. That there was a
slight bias (47%:53%) in the ratio of Kerry non-responders to
Bush non-responders"
............................................................
National Exit Poll
Kerry wins 50.8% of the total vote
Kerry wins 51.47% of the 2-party vote
Kerry Bush
2Pty 51.47% 48.53%
Exit 50.8 47.9
Recorded Vote Count
Kerry wins 48.28% of the total vote
Kerry wins 48.76% of the 2-party vote
2Pty 48.76% 51.24%
Actual 48.28 50.73
Bush needed 54.42% of refusers to match his 2-party vote.
But the Bush margin was just 53%-47% .
Kerry Bush
Responder 54.05% 51.47% 48.53%
Refuser 45.95% 45.58% 54.42%
Total 48.76% 51.24%
Did Bush get 53% of Refusers?
Then the refusal rate had to be 60.63% to match his vote.
But the refusal rate was 46%.
Kerry Bush
Responder 39.47% 51.47% 48.53%
Refuser 60.53% 47.00% 53.00%
Total 48.76% 51.24%
So which is correct?
Are we to assume that 54.42% of the 45.95% refusers were Bush
voters, while just 48.53% of those who responded were?
Or are we to assume that 60.53% refused and 53% were Bush
voters?