From: Kim C - PM
Date: 03 Nov 04 - 05:00 PM

The only way the exit polls could match the machine tallies is if EVERY single person who voted was polled EVERY single time. Nobody polled me.

And besides, we all know how biased "polls" can be.

KimC, you are wrong and this post displays stunning ignorance of statistics. Exit polls will match machine tallies if the same PERCENTAGE of people went for each candidate among the polling sample as among the voters. This is usually pretty easy to achieve. Because the order of voters, the willingness of voters to disclose their votes, and other things that affect the outcome of exit polls, are usually random, and other factors (number registered to each party) are known and can be controlled for, exit polls are usually quite accurate. This time I understand they were considerably less accurate than normal.

As for bias, typically in an exit poll if you were biased you would want to overreport the other candidate's numbers. This would make his supporters feel that it was less crucial to vote, while making your own candidate's supporters think it was MORE crucial to vote, suppressing his turnout and increasing yours. So if the exit polls show Bush losing and the election results show him winning, then two options are

1) the exit polls were jimmied for Bush.
2) the election results were jimmied for Bush.

The other options are

3) there was some unforseen regularity to the voting patterns eg. Bush voters voted later in the day on average)
4) by random chance, the sample was skewed toward Kerry voters.

As any pollster will tell you, #4 is the LEAST likely reason to have a discrepancy!