Exit polls showed that in NC Bush led Kerry by 4 percentage points (52-48) but final vote tallies showed a Bush margin of 12 points (56-44). I've been looking at how the Kerry and Bush numbers changed over election night and NC shows a large variation, especially early on.
First reports from cnn at 7:55 EST showed Bush with a large early advantage (28 percentage points, 64-36). This advantage stayed high for about the next 30 minutes, at least until 8:22pm (when 3.67% of the total vote was in). By 8:34pm, however, there had been a big shift to Kerry, so that the lead had dropped to 6 percentage points (53-47). This decreased margin continued at least through 8:43pm (when 12.2% of the vote was in). By 8:54pm (15.2% of the total vote reported) there had been a significant shift back in the favor of Bush, who now had a lead of 13 percentage points (56-43). This lead continued over pretty much the rest of the evening.
The same pattern shows up in the difference in actual votes (rather than percentages) early on election night. If the percentages had stayed constant over the night one would see a straight line through the origin (0,0). Instead the slope starts out high, then changes downward before staying steady as the rest of the votes come in, with the possible exception of a discontinuity at about 70% of the vote.
My question is, where does this spike come from? First, it's not unique to NC. A similar spike also shows up in SC and OH (also states with large actual vote - exit poll differences). In each case the spike shows up in the same form (i.e. up in Bush's favor, down in Kerry's favor, followed by a pro-Bush return), it shows up early (7:30-8:30pm or so, over by 9:00pm), and the bottom of the spike shows a level of Kerry support that's much greater than shown by the final tallies. At least in the cases of NC and SC the bottom of the spike closely approximates the exit poll numbers.
One possibility I've considered is that the spikes could be related to absentee/early voting. It's known in NC almost 30% of the total votes were absentee/early votes. Also, ignatzmouse showed that numbers for absentee/early votes in NC showed levels of support for Kerry and Bush that approximated the exit polls. So, could the spike in NC come from absentee/early voting? Absentee votes in NC are normally counted 5pm election day, so they could have been reported early. In any case, they had to have been reported by the end of the evening, otherwise the total numbers would have been much smaller. However, there are at least two problems with supposing the spike is from absentee/early votes. First, one would have to suppose that absentee/early votes for Bush began to be reported slightly before those for Kerry, in order for Bush to spike first. There doesn't seem to be any good reason to think that. Secondly, the spike that shows up in the vote count seems to be over significantly before all the absentee/early votes - there were almost 1 million - were accounted for. The spike ends between 12.2 and 15.2% of the total vote, i.e. between about 427,000 and 532,000 votes. On the other hand, if the spike isn't related to absentee/early votes, one would have to suppose they were fed in so gradually over election night that they barely caused a blip in the graph.
So, help please! Any ideas on this? Remember, too, that this isn't just NC, it shows up in other states as well.
Printer-friendly format
Email this thread to a friend
Bookmark this thread
(182 posts)
Send PM |
Profile |
Ignore
