Maybe the East was the Beast, but the ZOMBIES voted in the west.

[skids]

And the South was the HellMouth.

Another measure of the exit polls is how bad the "final adjusted" numbers are off in saying how many Bush2000 voters returned to the polls. That would suggest the worst areas were the West and South (as defined by the NEP.)

         %/Returning Bush2k voters/total              -3% mortality
               All voters                 real Bush2000          zombies      %zombies
EAST	36.00%	26128705	 9406333     9237792    8960658   445675	4.74%
SOUTH	49.00%	38076883	18657672    17580345   17052934  1604738	8.60%
MIDWEST	43.00%	30676093	13190719    13162586   12767708	  423011	3.21%
WEST	43.00%	26173241	11254493    10475279   10161020  1093473	9.72%

TIA's thread on exit poll disparity breakdown:

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=318693&mesg_id=318693

(The fine print: Alaska included in the West, though either Mitofski didn't or there was a doc error. Doesn't effect results much. Also "All voters" is just Bush + Kerry for now, until I get state-by-state ballots cast figures. But that will just make things worse.)

P.S. anyone got #of provisional ballots cast vs counted by state?

[Fly by night]

1. Can you realign the headings for your columns?

I'm not sure what heading goes with what column. Thanks.

[Land Shark]

2. Thank you, can you give a bit more detail?

Thankfully this is not as long as one of my posts, but I'm wondering if you can in a reply explain in more detail the definitions you used for returning Bush voters, zombies, etc. Thanks if you can, Skids!

[Bill Bored]

3. skids, while you're looking at this,

Check out the Dem/Rep party affiliation ratios in the question near the top of page 3. Notice how they changed from the second version to the third and final version on Nov. 3, particularly in the West. This is a huge weighting adjustment and it explains most of the vote shift to Bush. It's also just about the only weighting change that was made at this point in the reporting. All the other parameters are almost the same.

If there's any way to confirm these from poll records, etc. it would be very helpful. If the Dem turnout was suppressed (in REAL LIFE) to the point where it could affect these weightings, it could explain a lot.

I just can't spend too much time on these exit polls because I think there's more important work to be done (not that there's anything wrong with studying exit polls).

[skids]

4. OK, another look, as requested:

First, I'll flip the chart so the headers are less of a problem (though the first looks fine to me and the "pre" block should display correctly on any browser.)

                                        EAST    SOUTH  MIDWEST   WEST
% Returning Bush voters (NEP)             36%      49%      43%      43%
Total voters in 2004 (Bush + Kerry)  26128705 38076883 30676093 26173241
Number of Returning Bush voters       9406333 18657672 13190719 11254493
Number of votes Bush got in 2000      9237792 17580345 13162586 10475279
-3% mortality                         8960658 17052934 12767708 10161020
Number of zombie voters                445675  1604738   423011  1093473
Percent of zombies (NEP poll "error")   4.74%    8.60%    3.21%    9.72%  

OK, now as to what is going on here: NEP adjusted their poll without bothering to check how many people actually voted for Bush in 2000. As a result, their numbers are way out of whack. In fact, they are quite outside the MOE for this question, even assuming 100% of living Bush 2000 voters voted in 2004, which is a stretch in and of itself.

As to the numbers from previous NEP weightings. I don't know if the regional breakdown was included in the earlier weightings. I do know that if you run the full national numbers from one earlier weighting, you don't get nearly as many zombies (in fact, you have to subtract mortality to get any, whereas here, there are zombies even before you do) and I believe you end up within the MOE on this question.

The implications are unclear. The final weighting adjustment was drastic, yes, but the weighting problems might have started earlier than that.

I generally tend to avoid party affiliation questions because they don't have as much bearing on presidential vote as people seem to think. There are plenty of cross-voters and ticket splitters in both directions (one county I analysed in Ohio had a bunch of people voting Kerry plus a Republican down-ticket, so the opposite of "dixiecrats" do exist.) The end result could be anywhere. Anyway the suddenness of the late weighting can be shown just as well through this question.

As far as "better things to do," me too. But noone seems to want to adopt this seriously, and it is quite compelling to go to the CNN website, see numbers that are currently still posted, do grade school math, and produce results that are quite perturbing. After reading an article on KoS recently I was astounded at how many people on both sides of the fence over there are just not up to speed. So don't underestimate the tactical value of a can full of worms at this juncture. We aren't losing people, but they are getting tired, and some late "converts" could charge things back up a little.



(EDIT: typo)

[jkd]

8. Some questions

How many who were too young to vote or didn't vote in 2000, said they voted for Bush in the 2004 exit poll? How about population growth in the West and South in those four years? Does Kerry also have Zombies? Skids, keep up the good work. Your looking in the right places. I wish I had the time and the ability to the do the same kinds of investigations. I don't mean these questions as criticism. They are just a couple of explanations that came to mind.

[skids]

9. Answers.

The people who didn't vote in 2000 are beside the point, as these are people who said they voted. Neither does population growth matter. New voters are in their own category in the NEP poll.

Migration within the US isn't big enough to matter, and since all regions have zombies, it couldn't explain it anyway. If you want to see the numbers:

http://www.cnn.com/ELECTION/2004/pages/results/states/US/P/00/epolls.0.html#1064

Gore/Kerry's numbers (I don't have these worked out by region, just nationally.)

37% of 122188645 makes 45209799 people (statistically projected) who voted for Gore returned to the polls in 2004.

Gore got 50999897 votes. Minus 3% mortality, that's 49469900.

So no, there are no Kerry zombie voters. (The 122188645 is all presidential ballots cast, not just Bush + Kerry. I would use
that figure in the regional breakdown, as I said above, but first I need it by state.)

For Nader2000 and all other 3rd-party, there are zombies (4.6% nationally). That could represent an over-weighting of certain voters in that group, but such might be explainable as error due "binomial effect." So we can't really call foul there. (Plus 3rd party voters tend to be younger and the 3% mortality might not apply there. There are no zombies until the mortality is subtracted, unlike with Bush's figures.)

There is of course no way to do it for the "didn't vote in 2000" category, so we don't know what went on with the weighting there. There were 14.2% more votes in 2004 than in 2000. NEP says 17% of 2004 voters didn't vote in 2000. I could point out that if everyone who voted in 2000 came back in 2004, except for the dead people, the 3% mortality would require a bit under 3% new voters to make up for the dead people. And 14.2% plus a little under 3% is 17%. But that would be disengenous.

[Bill Bored]

10. You wrote,

"There were 14.2% more votes in 2004 than in 2000. NEP says 17% of 2004 voters didn't vote in 2000."

If these 2 statements are correct, that's 17% of 114.2% that didn't vote in 2000. To get the 2000 voters who voted in 2004: .83 * 1.142 = .94786
I.e., 100% of the 2000 voters, less the dead ones, etc. = 94.786% of the 2000 voters that voted in 2004. That's not impossible is it?

[skids]

11. It would be close to a 100% return rate

How likely that is, I'll leave up to professional election scholars.

My mortality rates are also estimates (and could easily vary between
the two parties, given that Bush scored higher among the elderly.)
Mortality is tedious to nail down these days because of the baby boom bulge and correlated demographic factors, so I attempted to err on the low side.

So to say it would work out to 3% to replace the dead voters
and about 2% to replace people who voted in 2000 and skipped 2004
is just an off-the-cuff estimate. Nothing impossible to see in this category, though, you are correct.

Only the Bush2000 figures are so far out of whack that they defy explanation.

[jkd]

12. Do these zombies suggest election fraud?

If the respondents were correct about their recollections from 2000, then 2,000,000 more people voted for Bush than his official vote. They were either mistaken about how they voted in 2000, or the weightings created this impossibility. With consideration for mortality rates, this would suggest that the weightings created about a 3% advantage for Bush or over three million votes in his favor.

Skids, are the other categories also inflated by the same factor? Have you observed any other contradictions in these exit poll data as a result of these last-minute weightings?

[skids]

13. Well, something isn't right...

...something on the magnitude of 3 million, yes. How many of those count double (vote switching versus simply not counting dem votes or stuffing) I can't even begin to speculate.

Unfortunately I haven't found another question on the exit poll which has hard, real-world numbers to back it up. Some hey can be made by looking at other polls in this respect. For example, the results from the Pew "religion and voting" survey would suggest conservative christians and orthodox jews were over-weighted. But it isn't a hard number like the vote counts from 2000.

Registrations and population ethnicity generally can't be used because they, unlike the 2000 vote tallies, always come up in the green due to the fact that not everyone who can vote, does.

I haven't been through each and every question to look for other hard numbers to compare the results with. If anyone comes up with an idea I'd love to hear it.

[Bill Bored]

14. I would point out that the 2000-voting question showed

that among the respondents, Bush got more votes than Gore.

[skids]

15. That wouldn't really be a problem...

...if it weren't for the fact that the survey seems to indicate close to 100% of people who voted returned to the polls in 2004. It would just mean there was a difference between the number of people on each side that decided to vote again.

As it is, how does it go? "Uh, Houston..."

[jkd]

16. Errors in exit polls

The accuracy of the data are no better than those who fill out the exit poll. Some who didn't vote four years ago will remember that they did or may just indicate who they would have voted for had they voted. Maybe that makes them feel better about not voting in 2000.

I see forms everyday that are filled out by well meaning people. They make lots of mistakes. They read over the questions too quickly to give a proper answer. I don't think that accounts for three million voters, though.

[Bill Bored]

5. All I can say is look at the original poll data

Party Affil was answered by nearly all respondents in all 3 versions of all 4 regionals. See the PDFs at scoop.co.nz. I don't this question was asked in the state polls though.

What do you mean no one takes this seriously? There have been lots of threads on this one. There are still lots of exit poll followers here!

I think it was the inauguration that reduced the post count on this forum, that's all. A lot of folks gave up after that.

[skids]

6. Re: taking it seriously.

Here, yes. But this forum is only a thin slice of even the liberal blogspace where this issue is being discussed. And there, no, there are still plenty of cooincidence theorists, yes-on-ohio-but-look-at-the-popular-vote people, would-have-required-cooperation-from-dems people, and I-know-there-was-fraud people who really don't have any grasp of the facts and figures we kick around in our fine forum.

I posted some stats on how many hits articles get in this forum a few days back. There are much more popular sites where most of us should be spending an hour each day, instead of obsessively reloading the page here.

[bleever]

7. Rec'd for Greatest.

This is great work, and this and TIA's stuff is the plain old stick that just has to keep being a nagging lump under anyone who wants to sleep peacefully over this.

Thus, kicked.

:thumbsup: