Not enough evidence in the exit polls

These were the exit poll numbers that a friend emailed me at 2 p.m. on Election Day:

Kerry is first number in all cases
Florida: 52-48
OH: 52-47
Michigan: 51-48
PA: 58-42
IA: 50-48
WI: 53-47
Minnesota: 57-42
NH: 58-41
ME: 55-44
NM: 49-49
NV: 48-49
CO: 49-50
AR: 45-54
NC: 47-53

How did the actual results for these states end up?

With Kerry the first number again, and with the difference between the exit polls and the results:
Florida: 47-52 Difference: 4/5 points
Ohio: 48.5-51 Difference: 3.5/4 points
Michigan: 51-48 - Difference: None. ACCURATE
PA: 51 - 48.5 - Difference: 6.5/7 points
IA: 49.2-50.1 - Difference: 2 points. ACCURATE
WI: 49.8-49.4 - Difference: 3-4 points
Minn: 51.1-47.6 - Difference: 5-7 points
NH: 50.3-49 - Difference: 8 points
Maine: 53.4-44.6 - Difference: 0-1 points. ACCURATE
NM: 48.9-50 - Difference: 0-1 points. ACCURATE
NV: 47.9-50.5 - Difference: 0-1.5 points ACCURATE
CO: 46.8-52 - Difference: 2 points. ACCURATE
AR: 44.5-54.3 - Difference: 0-.5 points ACCURATE
NC: 43.6-56.1 - Difference: 3-3.5 points


Well, let's use a different set of Exit Poll numbers. Here are the ones that Jack Shafer posted on Slate at 4:20 Pacific Time. These are later numbers.

Kerry first number
Florida 51-49 Difference: 3/4 points
Ohio 51-49 Difference: 2/2.5 points
Michigan 52-46 Difference: 1/2 points
Pennsylvania 53-46 Difference: 2/2.5 points
Iowa 50-49 Difference: 1 point
Wisconsin 51-48 Difference: 1/1.5 points
Minnesota 52-46 Difference: 1/1.5 points
New Hampshire 54-44 Difference: 4-5 points
New Mexico 50-48 Difference: 2 points
Colorado 49-50 Difference: 3 points
Arkansas 45-54 Difference: 0-.5 points ACCURATE
Missouri 47-52 Difference (46.1-53.4): 1 point ACCURATE
New York 62-36 Difference (57.8-40.5): 4.5 points
Nevada 49-48 Difference: 2.5 points
New Jersey 54-44 Difference (52.7-46.5): 1.5-2.5 points
West Virginia 45-54 Difference (43.2-56.1): 2 points

Well, by the late afternoon, the exit polls seemed to have corrected themselves -- albeit with Ohio still having a sizable difference in end result from the exit poll. Even if the exit polls still had Kerry winning Florida, it was still within 2 points -- an acceptable margin of error.

I'm not sure what explains the enormous shift in the exit polls throughout the day, especially given that, with the obvious exceptions of Pennsylvania, Minnesota, and New Hampshire's gigantic splits, most of the early day exit polls matched the most recent state by state polls that Zogby et al had compiled. Can the shift in the exit poll numbers be explained by Mickey Kaus' theory that the angry Democrats voted earlier in the day, and were evened out by the afternoon voting of conservatives? I'm not sure.

I have still not heard an explanation for how out of whack the Pennsylvania and Minnesota numbers were earlier in the day. But with Ohio and Florida so close -- the exit polls might have predicted Kerry the winner in some states that he lost, but by the 4 pm numbers, his lead in the exit polls from those states were slim.

Unfortunately, the exit polls used in this chart making the rounds would seem to be ones from earlier in the day, rather than the more accurate late afternoon ones.

I don't think the exit polls should be seen or used as evidence for a larger set of irregularities, though I firmly believe there were irregularities in Ohio and Florida that have not been addressed. (And should be. And hopefully, will be.)

But given how close Florida and Ohio were in both the exit polls and the results, I also don't think a lack of evidence in exit polls of irregularities should not be used as any kind of evidence that there weren't irregularities. A little tinkering can go a long way, and does not need to result in 5-10 point shifts.

Update; Dick Morris claims that “Exit polls are almost never wrong...So reliable are the surveys that actually tap voters as they leave the polling places that they are used as guides to the relative honesty of elections in Third World countries. … To screw up one exit poll is unheard of. To miss six of them is incredible. It boggles the imagination how pollsters could be that incompetent and invites speculation that more than honest error was at play here.” But instead of suggesting that this indicates there was some funny business in the actual voting, Morris instead suggests that the exit polls were fixed. Well. That's another way of looking at it.