Bush probabilities for various Party-ID weights in the National Exit Poll

[TruthIsAll]

Bush Final Percentage of the popular vote was 50.73%.

Let's calculate the probability of Bush winning 50.73% of the
popular vote, assuming various Party ID assumptions applied to
the voting percentages, based on the NATIONAL EXIT POLL 						

THIS IS THE ACTUAL PARTY-ID BREAKDOWN MIX FROM THE WP NEP
SITE.

38% DEMS, 36% REPUBS:
		HORIZONTAL			WEIGHTED		
PARTY ID	Mix	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	38%	9%	90%	1%	3.4%	34.2%	0.4%
Republican 	36%	92%	7%	0%	33.1%	2.5%	0.0%
Independent	26%	45%	52%	2%	11.7%	13.5%	0.5%
	100%				      48.24%	50.24%	0.90%
			Probability	5.30098E-07	or  1 in	1,886,445

****************************************************************
SENSITIVITY ANALYSIS:
PARTY-ID WEIGHT MIX AND CORRESPONDING PROBABILITY OF BUSH
GAINING FROM HIS EXIT POLL PERCENTAGE TO THE FINAL VOTE.

38% DEMS, 35% REPUBS:
		HORIZONTAL			WEIGHTED		
PARTY ID	Mix	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	38%	9%	90%	1%	3.4%	34.2%	0.4%
Republican 	35%	92%	7%	0%	32.2%	2.5%	0.0%
Independent	27%	45%	52%	2%	12.2%	14.0%	0.5%
	100%				       47.77%	50.69%	0.92%
			Probability	3.29447E-09	or  1 in	303,538,508

37% DEMS, 36% REPUBS:
PARTY ID	Mix	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	37%	9%	90%	1%	3.3%	33.3%	0.4%
Republican 	36%	92%	7%	0%	33.1%	2.5%	0.0%
Independent	27%	45%	52%	2%	12.2%	14.0%	0.5%
	100%				       48.60%	49.86%	0.91%
			Probability	1.49214E-05	or  1 in	67,018

36% DEMS, 38% REPUBS:
PARTY ID	Mix	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	36%	9%	90%	1%	3.2%	32.4%	0.4%
Republican 	38%	92%	7%	0%	35.0%	2.7%	0.0%
Independent	26%	45%	52%	2%	11.7%	13.5%	0.5%
	100%				      49.90%	48.58%	0.88%
			Probability	0.051889781	or  1 in	19

37% DEMS, 37% REPUBS
PARTY ID	Mix	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	37%	9%	90%	1%	3.3%	33.3%	0.4%
Republican 	37%	92%	7%	0%	34.0%	2.6%	0.0%
Independent	26%	45%	52%	2%	11.7%	13.5%	0.5%
	100%				      49.07%	49.41%	0.89%
			Probability	0.000569824	or  1 in	1,755

[Goldeneye]

1. So how did Bush win?

TIA, what percentage dems vs. reps (assuming ind. stay the same again) would be necessary for Bush to rake in 50.73% of the vote? I'm sure its been given before, but I can't remember...

[TruthIsAll]

5. For Bush to get 50.73%, it would have to be 35.0% Dem/39.0% Repub

EXACTLY 35.0/39.0

IS THIS WHAT THEY ARE GOING TO CLAIM?

I RECALL READING THAT DEMS OUTVOTE REPUBS BY 39-35%.


PARTY ID	MIX	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	35.0%	9%	90%	1%	3.2%	31.5%	0.4%
Republican 	39.0%	92%	7%	0%	35.9%	2.7%	0.0%
Independent	26.0%	45%	52%	2%	11.7%	13.5%	0.5%
	100%				      50.73%	47.75%	0.87%

[Goldeneye]

8. Cheaters I tell ya.

Good lord...what will it take to convince people?

[Bill Bored]

2. Is something wrong with the next to the last one?

It should be the same as the first one, shouldn't it?

In any case, the last one, which was what Mitofsky came up with, 37/37, shows almost a dead heat and it's within the MOE, right?

[Goldeneye]

3. They would be different

because more independents broke for Kerry, and because some cons voted for Kerry and some dems voted for Bush, but not in eqaul numbers.

[TruthIsAll]

4. Nothing wrong. Roundoff. The 37/37 mix is OUTSIDE the 1% MOE..

37/37/26:

1.67% deviation = 50.73- 49.07 = outside MOE.
The odds are 1 in 1,755. A long shot.

36/38/26:
 0.83% = 50.73% - 49.90%  is inside the MOE.
The odds are 1 in 19. A long shot at the track.

Bottom Line.

Kerry wins, even assuming 38Repub/36 Dem weights, which is
totally unrealistic.

I believe the actual weights are more like 39Dem/35 Repub.
I believe Kerry won by 51-48.

PARTY ID	MIX	Bush	Kerry	Nader	Bush	Kerry	Nader
Democrat 	39%	9%	90%	1%	3.5%	35.1%	0.4%
Republican 	35%	92%	7%	0%	32.2%	2.5%	0.0%
Independent	26%	45%	52%	2%	11.7%	13.5%	0.5%
	100%				      47.41%	51.07%	0.91%

Probability =1-NORMDIST(0.5073,0.4741,0.01/1.96,TRUE)
Probability = 3.84717E-11	

or ********** 1 in 25,993,152,685 ********

[Bill Bored]

6. My mistake

I thought they were both 38/36.

But as far 37/37, the actual vote count is outside the MOE, as you say, but the election result (popular vote) is within the MOE:
49.07% B/49.41% K

But, if you throw in some ballot spoilage of say 2% nationally, which favors Bush, it could account for the actual margin.

In any case, with 37/37 Bush is within the MOE of winning the popular vote, and so is Kerry.

And my point is that this is all because of this one simple weighting change.

[TruthIsAll]

7. But we know that 37/37 is totally bogus. It has been 39/35 Dem

over the last few elections.

Look at the odds of the percentage move: 1 in 1,755

[Kansas Wyatt]

10. Apparently some new scientific information has come into the MSM...

It turns out that there was absolutely nothing to worry about all along, and it was actually the fundie voters that took Shrub over the top.

According to Scientific Government Data just released, the fundies have been multiplying themselves by separating into two individuals or voters. At this rate dividing themselves repeatedly for the last four years leading up to the election, this continuous multiplication is actually how Shrub seems to have defied statistics and mysteriously gained millions of new voters out of thin air.

* At least it's a better excuse than anything they have thrown out so far, isn't it?

Since that has been settled, now we can get on with another terrorist attack and war, throwing Social Security in the stock market, permanent tax elimination for the rich, and making American citizens officially subjects of the Bush Dynasty Government.

[Bill Bored]

12. Well, I think this is something that should be checked out

If it wasn't really 37/37, then you have a real counting anomaly. And perhaps, unlike the votes cast on secret ballots with secret source codes with no permanent record, this can be verified somehow.

We know there was suppression of Dem voting. Even Democratic registration! If this didn't happen on the Repub side, then they could have equalized the turnout.

[rdmccur]

9. That is, in fact,

about what the exit polls were indicating (what you believe).

[davidgmills]

11. Give em Hell TIA

If Mytofsky is listening, and I would bet on it, he's got to be worried sick.

Damn we've sure got a lot of Republicans in this country. Multiplying like cockroaches. That's what freepers were telling me before 11/2.

[flintdem]

13. Weighting is not that simple

The 200 VNS Exit poll weight variable ranges from .111 to 8.348 with nearly 1000 values in between. How ever exit polls are weighted, they use multiple combinations of demographic, region and party id variables put into one weight variable to produce the outcome.

The 1996 VNS is similar with .15 to 5.18 weighting with several hundred values in between.

[TruthIsAll]

14. You can always calc the weights needed to produce a result.

Goal seeking. What-if.

Takes time. But surely not 10 months.

[flintdem]

15. Past of the time is validating the data

Tests are run to make sure that the data is valid down to the precinct level. Handwriting of individual respondent's forms are compared to make sure that the interviewer didn't just fill a few out themselves (and skip doing their job- which has happened in some polls). It is not highly unusual for a poll to drop a few respondents results as being suspect.

They also re-check the data entry against the original forms to correct for coding errors (which always happen). Keep in mind that on election night large numbers of people are doing the coding. I wouldn't be surprised to find out that this job is left one or a handful of people (it's not priority data anymore). Your talking about checking maybe 70 variables for 95,000 respondents (including state polls) or 6,650,00 individual entries!!! Remember- they don't make any money off of cleaning this data up and putting it in the public domain. In the past no one has been interested in it after the election except academics.

There is also a very small possibility that they have had contracts with academics who get first crack at the data for writing articles and publishing before anyone else gets it. But it would have to be a very short lead time.

[Bill Bored]

16. But the point is that

with one very simple weighting change, the election becomes a statistical dead heat, and this parameter (party affiliation) can probably be verified independent of the exit polls.