National Exit Poll Analysis: Using FACTUAL Historic Data

[TruthIsAll]

2004 Election Analysis Model					

TOTAL VOTER TURNOUT					
122.26 million voted in 2004				
104.78 million voted in 2000				
for a
17.48mm net increase in voters				
					
BUSH 2000 VOTER TURNOUT
50.456mm voted for Bush in 2000				
1.766	 died (3.5% approx.)				
0.487	 did not vote (1% approx)				
So the maximum number of Bush voters who could have voted in
2004 is:	
48.203mm (39.43% of 122.26mm)				
					
GORE 2000 VOTER TURNOUT					
50.999mm voted for Gore in 2000				
1.785	 died (3.5% approx.)				
0.492	did not vote (1% approx.)				
So the maximum number of Gore voters who could have voted in
2004 is:
48.722mm (39.85% of 122.26mm)				
					
NADER 2000 VOTER TURNOUT					
3.322mm voted for Nader/other in 2000				
0.116	died (3.5% approx.)				
0.032	did not vote (1% approx.)				
So the maximum number of Nader/othervoters who could have
voted in 2004 is:
3.174mm (2.60% of 122.26mm)				
					
Therefore, the MAXIMUM POSSIBLE turnout in 2004 of those who
voted in 2000 is:		
Bush	48.203mm	39.43%			
Gore	48.722	      39.85%			
Other	 3.174	       2.60%			

TOTAL	100.099mm    81.87%			
					
So the MINIMUM number of NEW voters had to be:			
NEW	22.161mm     18.13%			
					

NATIONAL EXIT POLL TIMELINE:
Poll 1: Posted on CNN 11/02 @ 7:38pm (11027 respondents)
Poll 2: Posted on WP 11/03 @ 12:22am (13047 respondents)
Poll 3: Final posted on CNN 11/03 @ 2:05pm (13660 respondents)
		
KERRY SHARE OF NEW VOTERS:
Poll	Kerry% Votes	Bush%	Votes	Kerry margin (mm)
1	59%	13.08	39%	8.64	4.43
2	57%	12.63	41%	9.09	3.55
3	54%	11.97	45%	9.97	1.99
					
					
CALCULATING THE VOTES:
Applying the FACTUAL historic weightings computed above to the
Exit Poll timeline, from the First (11027 respondents) to the
Second (13047) to the Final (13660), we get: 

Exit Poll 1:
Kerry winning margin: 7.67 million votes	
					
Voted	2000	Mix	Bush	Kerry	Nader
22.161	None	18.13%	39%	59%	2%
48.722	Gore	39.85%	8%	91%	1%
48.203	Bush	39.43%	90%	9%	1%
3.174	Other	2.60%	13%	71%	16%
					
122.26	Total	100.00%	46.08%	52.35%	1.57%
		122.26	56.34	64.00	1.92
					
Exit Poll 2:
Kerry winning margin: 6.78 million votes	
					
Voted	2000	Mix	Bush	Kerry	Nader
22.161	None	18.13%	41%	57%	2%
48.722	Gore	39.85%	8%	91%	1%
48.203	Bush	39.43%	90%	9%	1%
3.174	Other	2.60%	13%	71%	16%
					
122.26	Total	100.00%	46.44%	51.99%	1.57%
		122.26	56.78	63.56	1.92

A PROBABILITY CALCULATION
The margin of error for a 3168 randomly-selected sample is
+/- 1.78%. Assuming this MOE, the probability of Bush having
46.44% in the poll and ending up with 50.73% of the vote is:
1 in 984,473
				

Exit Poll 3:
Kerry's winning margin is 3.29 million votes.	
					
Voted	2000	Mix	Bush	Kerry	Nader
22.161	None	18.13%	45%	54%	1%
48.722	Gore	39.85%	10%	90%	0%
48.203	Bush	39.43%	91%	9%	0%
3.174	Other	2.60%	13%	71%	16%

122.26	Total	100.0%	48.36%	51.05%	0.60%
		122.26	59.12	62.41	0.73


THE 22 MILLION NEW VOTERS:
Bush needed 64.5% of the 22.161mm NEW voters to gain his 3.0
mm margin. That means Bush had to win 14.294 mm(64.5%) New
voters and Kerry only 7.867 mm (34.5%). 

The FINAL Exit Poll (13,660 respondents), which was matched to
the vote, gave Bush ONLY 45% of NEW voters. 

The first 11,027 respondents gave him 39%.
The first 13,047 respondents gave him 41%.
					
Voted	2000	Mix	Bush	Kerry	Nader
22.161	None	18.13%	64.5%	35.5%	0%
48.722	Gore	39.85%	8%	91%	1%
48.203	Bush	39.43%	90%	9%	1%
3.174	Other	2.60%	13%	71%	16%
					
122.26	Total	100.00%	50.70%	48.09%	1.21%
		122.26	61.99	58.80	1.48

To believe Bush won 64.5% of NEW voters, then one must also
believe that 5 million (41%) of all NEW voters who said they
voted for Kerry must have LIED to the pollsters. 

[Rex_Goodheart]

1. My immediate reaction to that analysis

is that the exit poll number which shows that 9% of Bush voters in 2000 switched over to Kerry is probably overstated. I can count on one hand the number of ex-Bush voters I know who switched over to Kerry.

If there were some other pre-election poll that reported a similar rate of dissatifaction this latest well-done analyses of yours would carry a lot more weight, IMHO.

[TruthIsAll]

2. But what about the 8-10% of Gore voters who switched to Bush?

Most anecdotal evidence indicates that there were more Bush voters who switched to Kerry than Gore voters who switched to Bush.

The interesting part of this is that Kerry wins even when the numbers are overweighted heavily in Bush's favor to match the vote (see Final Exit Poll).

Two examples:
1.The 41% weighting of returning Bush 2000 voters. Can't be. Bush has to be at 39% tops.

2.The 54%-45% spread of Kerry support among NEW voters. That is much lower than the 59-41% spread in poll 2 and the 59-39% in poll 1.

[Rex_Goodheart]

3. Well, I don't think anecdotal evidence suffices

when you're trying to build a case, but....

I agree with your analysis. I do not for a moment believe that more new voters voted for Bush than for Kerry. Nor do I believe that more Gore voters switched to Bush than Bush voters switched away, if at all.

So, where in the world did Bush get his majority? I believe that if that 9% who left Bush could be corroborated the game is over... the case for corruption has been mathematically proven.

[TruthIsAll]

5. I am not using anecdotal evidence at all. Strictly the exit poll

percentages, with adjustments to the weightings - from the mathematically impossible 43/37% to the very plausible 39.43/39.85%.

There is no bias in these weights, or polling margin of error, for they are not based on how people said they voted in 2000, but rather on the historic fact.

Why not use the 2000 voting results along with the mortality rates that we know to be true as a maximum constraint on the weightings?

And just work from there?

[libertypirate]

23. Depends what part of the country....

I can't find a republican in CA who doesn't watch FOX and voted for Bush.

[Anarcho-Socialist]

4. I'm very appreciative of all the statistical work you do...

..and it does seem that there is strong circumstantial evidence in favour of something unsavoury (fraud) occuring in the voting process.

[Rex_Goodheart]

6. TruthIsAll, suppose Bush voters...

returned to the polls in 2004, but Democrats stayed away?

In other words, even if 9% of each switched parties, Bush could have gained his 3.2 million vote majority if 100% of his voters returned while only 93% of Democrats did... wouldn't this also partially explain the 43%/37% mystery, i.e why more people said they had voted for Bush in 2000?

Edited to add the following:

Actually, though, it seems that this election needed near 100% of both parties to numerically make sense. The Mitofsky exit poll says that 17% of voters were new, which is near 21 million... So I guess we need both sets of 48 million to reach 122 million total.

[TruthIsAll]

7. SENSITIVITY ANALYSIS: THE KERRY NATIONAL VOTE MARGIN OF VICTORY...

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x350225

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x350526

[Rex_Goodheart]

15. Trying to come up with something plausible...

that would match the official election returns.

In order to make it work, I had to:

- raise new voters to 25 million (significantly above the 17% Mitofsky reported)
- limit Bush's abandonment rate to 6%
- raise the Gore/Kerry abondmentment rate to 11%
- enforce 100% reparticipation for year 2000 Bush voters
- limit Kerry voters reparticipators to somewhat less than 100%

Note: this NOT not what I think happened...

<img src=>

[TruthIsAll]

16. So, according to your scenario, for Bush to win by 3 million

1. There had to be 25 million NEW voters, not 22 million.
That's a 13.6% polling error.

The MoE for a 3168 sample-size is 1.78%.

The probability of the discrepancy is:
= 1- NORMDIST(0.565,0.43,0.0178/1.96,TRUE)
or less than .00000 00000 00000 00000 00000 000001

2. Kerry only won 13/25 or 52% of New voters. Bush had 48%.

The Final Exit Poll of 13660 respondents, which was matched to the vote and had impossible 43/37% weights, had it as 54-45%.

The 3% difference was beyond the 1.78% MoE.

The probability of that discrepancy is:
= 1-NORMDIST(0.51,0.48, 0.0178/1.96,TRUE) = 0.000478
or 1 in 2093

3. 98.5% of Bush 2000 voters (those still alive, of course) turned out, while only 93.4% of Gore voters did.

I guess 1 out of 14 Gore voters, who felt the election was stolen from them by the Supreme Court, were not motivated to vote this time and decided to stay home, rather than vote for Bush.

4. Five of 46 Gore voters (10.8%) voted for Bush to reward him for his SCOTUS-led selection in 2000.

5. Three of 48 Bush 2000 voters voted for Kerry this time (sounds reasonable).

I guess 1 out of 5 is not too bad.

[Rex_Goodheart]

17. It's not really my scenario..

I believe that if the opportunity is there for corruption, then corrupted it will be. And corrupted it was.

Can we assume that people who didn't vote in 2000 but were eligible might be a little embarrassed to tell that to an election pollster? Isn't there a general stigma attached to not voting? And, if so, wouldn't that mean that the 17% Mitofksy reported as new voters was low?

If that 43%-37% percentage were recognized as really 40%-37%-3% embarrassed so they said Bush, then the scenario I constructed makes good sense.

But isn't it funny how one has to assume so much unsupportable stuff just to make Bush's victory make sense?

[liam_laddie]

18. NEP poll answer method

SFAIK, based on E/M's letter to the BoE Directors, the polling
was to be NONverbal. In my book, this means the pollster asks who wishes to respond, (guideline is every 5th/6th voter, I think,) hands a clipboard with question form to voter, voter fills in answers, hands board back to pollster. Forms used were either one- or two-pages, depending on purpose. There wasn't supposed to be any "face-to-face" questioning. Seems to reduce the embarrassment factor re: misleading answers.
The completion rate might be affected by those voting early, hurrying on way to work. Any possibility?

There is a non-response rate of around 50%, and some non-completion factor as well. In any case, I'm studying one precinct, with a high red-shift, in some detail... Of (431) votes cast at the poll, (31) exit poll samples were recorded by E/M. This would indicate about every 7th voter here was asked, and about half of those were completed. I believe the target was about 40-45 per precinct. I mention this to assist those who don't have a "picture" of the polling technique, as I understand it. HTH.

[TruthIsAll]

8. One other salient fact.

Mitofsky changed the weights from 41/39% to 43/37%, not the respondents.

The first 13047 respondents had it at 41/39%

41% of 13047 = 5349
43% of 13660 = 5873

Do you believe that of the final 660 respondents, 524 (79%) claimed to be Bush voters, when only 41% of the first 13047 did?

[Rex_Goodheart]

11. Actually, no... the last 613 respondents

could not have changed the rate from 41% to 43%... however, that view is not a correct representation of the change.

What happened, I think, is that all responses are weighted to the returns, and the chips fell where they may. So, just as Kerry ended up losing in the final poll, even when those last 613 respondents could not have swung the poll, so did every other percentage in the poll.

[TruthIsAll]

12. Now you've got it. That's what I have been saying all along.

Q.E.D.

[dzika]

9. Is that last statement worded correctly?

To believe Bush won 64.5% of NEW voters, then one must also
believe that 5 million (41%) of all NEW voters who said they
voted for Kerry must have LIED to the pollsters.

There couldn't have been 5 million voters that lied to pollsters if there were only 3168 responses. I understand what you mean, though.


Another question...

Do these calculations assume that the "official" total number of voters is correct and that one or more of the following must have happened:

1. Voters gave incorrect information to exit pollsters
2. Final exit poll weighting is incorrect
3. Final vote count (but not total votes) is incorrect

Are there more possibilities that I'm missing?

[TruthIsAll]

10. One at a time...

Yes, the wording is incorrect. It should read:

To believe that Bush won 64.5% of NEW voters, then one must also believe that 41% of all NEW voters who said they voted for Kerry must have LIED to the pollsters.

To put it another way:

If a) 41% (5349) of the first 13047 polled said they voted for Bush
and 39% (5088) said they voted for Gore and

b) 43% (5873) of the full 13660 said they voted for Bush
and 37% (5054) said the voted for Gore

then

c) of the final 660 respondents, 524 (79%) must have claimed to have been Bush 2000 voters and

d) 34 of the original 5088 Gore voters must have vanished and no others showed up among the final 660.

Of course, that is not plausible, yet we also know that changes in the weights were made to match the vote count and could not have been from sampling the final 660.

Come to think of it, why were the final 660 sampled, anyway?

You:
Another question...

Do these calculations assume that the "official" total number of voters is correct and that one or more of the following must have happened:

1. Voters gave incorrect information to exit pollsters
2. Final exit poll weighting is incorrect
3. Final vote count (but not total votes) is incorrect

Me:
Yes, without evidence to the contrary, we must assume that the official number of votes is correct for both 2000 and 2004.

If there were more than 122.26 million votes in 2004, then the weights would be even lower - for both Bush and Kerry. The number of NEW voters would be even higher. And that would also mean that NEW votes were discarded. Kerry won a significant majority of NEW voters.

If there were fewer than 122.26 million votes, then that would mean that votes were padded in the final count.

As to the other questions:

1. There is NO evidence of this.
2. YES, absolutely, it is mathematically impossible.
3. YES, if only impossible weights show that Bush is a winner, and ALL combinations of possible weights show that Kerry is the winner, then there can be just one conclusion.

You:
Are there more possibilities that I'm missing?

Me:
None come to mind at this time.

[Sancho]

13. Finally, this is what makes the most sense...

Without the RAW data, bandwagon effects and other excuses are impossible, as I've stated on other threads, because hundreds of polled voters representing millions of votes could not be so out of reality in such numbers without the pollsters noticing...so

1.) either the RAW support that Kerry won and we need a BIG investigation or 2.) the RAW data support that E-M are the most incompetent pollsters ever born

Occom's Razor: what is the most likely and simplest explanation without regard to morality - 1.) or 2.) ?

[Sancho]

14. Where is Q when we need him/her?

Here's the interesting thing! What if the number of voters is inaccurate because the tabulators were hacked and we really don't know the total vote in 2004 (or 2000) for that matter? Millions of votes may have been "added" or "subtracted"? The percentages don't make sense! again, how about the RAW data so we can conclude a BIG investigation is in order!

[dzika]

19. Looking at NEP's raw data

NEP's released raw data doesn't provide a way to determine the order in which responses were compiled. There is no obvious way to determine which raw responses were used in the 7:38pm or 12:22am polls.

Out of 3175 actual exit poll responders who were polled about their 2000 vote, 585 indicated that they did not vote in 2000.

Do the 585 actual responders represent a 17.48mm net increase in voters instead of 22mm?


Raw Vote2000 responders:

		Bush	Kerry	Nader	NoVote	Other	Total
		
	Total	1488	1638	20	7	22	3175
	
Bush 2000	1117	122	5	3	6	1253
Gore 2000	103	1108	5	0	3	1219
No Vote 2000	248	327	4	3	3	585
Other 2000	20	81	6	1	10	118

Weighted Vote2000 responders:

		Bush	Kerry	Nader	NoVote	Other	Total
		
	Total	1607	1523	15	9	14	3168
	
Bush 2000	1230	120	4	5	4	1363
Gore 2000	115	1042	3	0	2	1162
No Vote 2000	240	285	5	2	2	534
Other 2000	22	76	3	2	6	109

[dzika]

21. I'm still trying to understand...

How do the raw exit poll responses relate the actual vote counts in the 2000 and 2004 election?

Do the last 660 respondents (or the last ~165 vote2000 responders) contaminate the results in the final raw exit poll?

If the last 660 respondents have indeed contaminated the poll, then is it really possible to make assumptions based on the final poll or the raw responses that the final poll is based on?

[TruthIsAll]

22. Answers

You
How do the raw exit poll responses relate the actual vote counts in the 2000 and 2004 election?

Me
They are closer than the impossible final, but not perfect. 41% Bush/39% Gore is still too high for Bush.
His max is 39+%

You
Do the last 660 respondents (or the last ~165 vote2000 responders) contaminate the results in the final raw exit poll?

Me
You must forget about the final 660. They were just an artifice in the "adjustment" of the 13047. They are obviously not based on reality, if you consider Kerry LOST votes and Bush gained them all in going from 13047 to 13660.

You
If the last 660 respondents have indeed contaminated the poll, then is it really possible to make assumptions based on the final poll or the raw responses that the final poll is based on?

Me
Of course. It shows that the 13047 pristine poll was close to being right, and that the 13660 final which matched the vote was obviously wrong, as is proven by three bogus demographics:

PartyID: 37/37/26?
No way. It was 39/35/27 in the prior to elections.

How Voted 2000: 43/37?
Not possible. Bring it down to a still unbelievable 41/39, and Kerry wins. Bring it down to 39/40, Kerry wins in a landslide.

Gender: 51 Kerry/48 Bush?
Impossible. Gore got 54% of females.

[dzika]

24. raw data is not time stamped; impossible to tell which responses were last

The individual response data released by NEP doesn't provide for a way to determine which records were added to the poll last. The last 660 or whatever physical number of responses that correspond to the final 660 can't be determined based on the data released from NEP.

What is throwing off the final poll? Is it the final 660? You can see from the tables that I posted above that the weights are almost insignificant. In fact, in NEP's final set of raw data, Bush wins even when no weighting is applied to the sample.

I thought that you had determined earlier that the 660 final samples could not account for the difference between the final poll and the 12:21am poll.

I'm trying to understand the physical evidence here and I can't dream up a satisfactory explanation for how we got from the 12:21am poll to the final regardless of the 660 final samples.

Are the final 660 records in the raw data fabricated?
Were some of the 13047 records altered after 12:21am?

IF the final set of raw records match the final exit poll (and I think that they do)
AND if your calculations show that the final poll has impossible totals based on the 2000 vote
AND if the final 660 samples and weighting alone can't explain the difference between the 12:21am and final poll
THEN didn't there have to be changes to some of the 13047 raw samples
AND doesn't this suggest that the NEP has perpetrated a hoax beyond simply weighting the vote totals to match the vote count?

I know, it's a long question but it's also a complicated problem.

[TruthIsAll]

25. Long question, short answer

Draw your own conclusions:
Kerry went from 54% FEMALES in 13047 to 51% in 13660?
Kerry went from 57% NEW voters in 13047 to 54% in 13660?

You can also compare many other demographic category vote discrepancies.

No, it just wasn't the weights.
Changing them wasn't enough to match the vote count.
The percentages had to change as well.

Smells real bad.

[Kip Humphrey]

26. One of your strongest conclusions to date

What I hear you saying is: It was impossible to adjust the exit polls to reflect the reported results without altering/manufacturing the percentages. The demographic results which the media touted as the only reliable data in the exit polls - quoted religiously by Fox, CNN, and Republicans (remember the "values" BS?) made sufficient adjustment impossible because they then produce impossible results and thereby expose the falicy of the final numbers. Put another way,

A farmer says he has 6 lemons and 5 apples. Only problem is 5 fruits are yellow and 6 are red.

So, the farmer pulls out a paintbrush and paints one of the apples yellow. Now he has 6 yellow colered fruits and 5 red colored fruits.

The only problem is: 6 fruits are shaped like apples and 5 are shaped like lemons. So, the farmer pulls out a knife and carves the yellow colored apple into the shape of a lemon.

Now, the farmer has 6 yellow fruits shaped like lemons and 5 red fruits shaped like apples

The only problem is the yellow, lemon-shaped apple starts to rot and smell really bad!

[dzika]

20. kick. n/t