Reply #13: One more go. You proved one of their numbers had to be innaccurate.

[Brotherjohn]

The question, and dispute, is then "does this prove the entire election results are wrong?"

You:
You fail to see that the 41.26% is not dependent on how many were polled. It is just the fraction of the total number of Bush 2000 voters (50.456) divided by total votes (122.26) in 2004.

Me:
I do in fact see that the 41.26% is not dependent on how many were polled. That is exactly my point. You are using a total known number to question the results of a limited sample of the exit poll. You prove the maximum possible in reality. I accept that part. But exit polling is not reality. The exit poll isn't necessarily going to match it. It only proves that that individual exit poll number cannot be accurate. Our dispute is then in what THAT can further prove.

You:
So it is AN ABSOLUTE MAXIMUM.
IT IS IRREFUTABLE ARITHMETIC.
AND THAT IS THE BASIS FOR THE ANALYSIS.

Me:
It is the absolute maximum of the universe. It is not the absolute maximum this poll can yield. The poll should yield no more than 1% greater than 41.26% IF the sample is indeed random (based on MOE). But you are assuming the sample is random.

You:
Must I resort to basics?
No one polls 122 million people.
They poll a representative sample.

Me:
Duh. That's my point... except I didn't say "representative". They're not polling the entire 122M, so you can't say 41.26% is the absolute maximum possible. A poll sample can yield a higher, perhaps MUCH higher, number if there are problems with the sampling and methodology. If we knew for SURE they polled a perfectly representative sample (something easier said than done) you might be onto something.

You:
The accuracy of a poll has nothing to do with how many are in the universe (here 122 million) to be polled. I'm sure you will disagree, but you would be wrong.

Me:
Duh again. I didn't get all A's in them, but I took a number of statistics courses in grad school, have a doctorate, and use statistics in my everyday work. I know the basics.
My point was precisely the the one you subsequently make: that the accuracy depends on the sample polled (and that the sample has to be random). But you are using what is an unquestionably accurate calculation based on total numbers (some from 2000, some from 2004) to question a percentage in the exit poll. But the exit poll could very well be inaccurate.

Rather than continue to address every point, let me just say that I follow the logic throughout the rest of your post. But it seems what you have done is found a solid device/calculation to prove that the exit polling is faulty. YES, the MAXIMUM POSSIBLE percent of the 2004 voting electorate that voted for Bush in 2000 is 41.26%, based on your correct calculation. But this is only a part of the equation. It proves that this number in the exit polling is wrong.

But here I think is your major logical flaw: just because that number in (or even the entire results of) the exit polling is wrong, and just because the exit polling numbers match the final official results, that doesn't logically make the final official results wrong! The source of the problem with the exit polling numbers is their percentage of voters who voted for Bush in 2000, which you have shown is an impossibility.

But the official tally has nothing to do with this calculation. The official tally was not based on the exit polling. Who KNOWS what the hell kind of tinkering Edison-Mitofsky did to make their polls look like the final ones (to avoid embarrassment, to facilitate fraud, whatever...).

To answer a few final points in closing:

You:
You fail once again to see the obvious.
I will repeat.
Follow closely.
THE FINAL EXIT POLL IS WRONG - WE JUST PROVED IT.

Me:
You didn't prove the total numbers in the final poll were wrong. You proved that one of the numbers they used in calculating their total numbers was wrong. Other numbers could have been equally skewed to opposite effect to yield the final accurate result.

You:
THE FINAL EXIT POLL MATCHED TO THE RECORDED VOTE - WHICH BUSH WON.
IF THE FINAL EXIT POLL IS WRONG, AND IT WAS MATCHED TO ANOTHER NUMBER (THE RECORDED VOTE) THEN....
THE RECORDED VOTE MUST ALSO BE WRONG.

Me:
Correlation does not prove causation.

You:
I am sure that as you come to understand the logic, you will become convinced.

Me:
Sorry. I think your logic is flawed. You may have proven that one of the numbers within the exit polling is mathematically impossible. But that only proves anything from their sampling being non-random to their entire polling being intentionally skewed. It proves a problem with their polling, and perhaps their attempts to make it match the final results. But the fact that it matches the final results does not prove that the final results are likewise problematic.

NOW... IT VERY WELL COULD BE that their final exit poll numbers are flawed because they changed them (and you have caught an error in their changes) to reflect the "official" (doctored) tally which showed Bush winning, and that their initial exit poll numbers were more representative of the ACTUAL voting going on, and that massive fraud DID indeed occur. I just don't think this proves it.