Reply #22: Well, I already addressed these, prolifically, but....

[Febble]

FEBBLE:
I think the evidence suggests that he won the popular vote - I'm less sure about the EC vote, but I do indeed maintain that the early state exit polls and the 12:22am National Exit Poll (NEP), which Kerry won by 51-47%, are not convincing evidence of fraud.

TIA:
Ok, then where did Bush find 16 million new voters?
http://www.geocities.com/electionmodel/TruthIsAllFAQRes...

That is not a response to my point. I don't know the answer to your question, but I can suggest some places to look.

FEBBLE:
I have provided extensive statistical evidence to back up my claim.
TIA:
Show us the evidence.

http://inside.bard.edu/~lindeman/slides.html
http://journals.democraticunderground.com/Febble/3
http://journals.democraticunderground.com/Febble/7

Plus the finding I mentioned here:

http://www.democraticunderground.com/discuss/duboard.php?az=post&forum=203&topic_id=467048&mesg_id=467102

Plus the the strong correlations I found (as also reported in E-M evaluation document) between the magnitude of precinct level discrepancy and methodological variables, including the interviewing rate.

FEBBLE
as long as you discount the raw precinct-level questionnaires that have been archived at the Roper Center since January 2005, and were available for public download for more than a year, as well as the precinct-level vote shares for Ohio, which were "blurred" to prevent voter identification, and published in a paper by ESI. In addition, of course, state-level close-of-poll estimates based on a) raw, stratified exit poll data and b) raw stratified exit poll data weighted by pre-election polls have also been published, and publicly available since January 2005.

TIA
Why don't you download it for us?

Too late. You had a year to do it in for free, but it seems you didn't bother. I don't have a subscription. If you want one you will have to pay for one. I gave you the URL on numerous occasions.

FEBBLE:
All conclusions should be subjected to rigorous testing, and further disambiguation, and, to my knowledge (seeing as I did a fair bit of myself), were.


Tortured explanations for exit poll discrepancies include but are not limited to the following: Kerry voters were more likely to respond to exit pollsters; exit poll interviewers sought out Kerry voters; Bush voters lied or forgot who they voted for in 2000; polls are not true random samples; exit polls are not designed to expose fraud in the U.S. They point out that Democrats always do better in the polls than in the vote count because of this endemic bias.


None of which are "tortured" at all, and many of which are supported both by the 2004 data and by data from other elections.

TIA:
Explain why these assumptions are not "tortured".
Explain the rational for the derived Bush vote shares.

http://www.geocities.com/electionmodel/TruthIsAllFAQRes...

I have explained before, on countless occasions. I see no particular reason to do so yet again. But in short:

There are only two reasons why there should be an discrepancy between an exit poll and the count:

1. the votes were miscounted in favor of one candidate (i.e Bush, over all in 2004)
   
2. voters for one candidate participated in the poll at a higher rate than voters for the other (i.e. voters for Kerry, overall, in 2004).

Frankly, I suspect some of both, but the evidence strongly supports (2), only weakly supports (1), and actually contra-indicates (1) by DREs, or on a scale of millions. So that leaves (2). Why Kerry voters should have participated at a higher rate than Bush voters cannot, of course be determined. However, there is strong evidence that they did so. This evidence lies in the strong correlations between the magnitude of the discrepancy and methodological factors that would have tended to make departures from strict Nth voter sampling protocol more likely. It is also borne out by actual experimental evidence, in which exit poll discrepancies (in the redshift direction) have been associated with certain methodological factors. It is also borne out by an opinion poll conducted shortly before the 2006 election that indicated that Democrats were more likely to take part in an exit poll than Republicans. It is also borne out be evidence from Steve Freeman's exit poll (you'll have to google for that, I'm afraid, don't have the link to hand). It is also borne out by a study of exit poll discrepancy by Michael Butterworth (on data from 2006). There is massive evidence that the polls are not true random samples - this is evidenced by the non-response data apart from anything else. No pollster can assume a random sample. Your own pre-election polls data indicates that polls have a substantial amount of non-sampling error. There are vast text books written on the subject. There is also substantial evidence both from longitudinal panel data, and from past exit polls, that people tend to over-report having voted for the incumbent. There is also evidence to suggest that they over-report having voted at all.

TIA:
They never consider that in every election, a significant percentage of total votes cast are never counted and overwhelmingly Democratic.

FEBBLE:
This is simply untrue. Not only did "they" consider it, but "they" actively investigated whether this was a likely contributor to the exit poll discrepancy.

TIA:
It is untrue? Where has it been shown to be untrue?

Well, I'm part of "they" and I investigated it - for Warren Mitofsky, so he can be "they" as well. Yes, I considered it, and yes I reported my findings to Mitofsky who was interested in the result.

FEBBLE
And again, this is not simply untrue, but a lie. TIA knows it is untrue; in any case it clear from the Edison-Mitofsky report that at least one fraud hypothesis was actually tested. I myself tested more.

TIA:
Are you saying there was no fraud? I may be mistaken but I never lie. Show us some examples aof where you and Mitofsky considered fraud.

The Edison-Mitofsky evaluation reports findings from an investigation into the correlation between voting technology and exit poll discrepancy. I repeated this investigation, rather more rigorously, I think (see above). Also check out the swing-shift correlation above. That was a direct test of the hypothesis that fraud was responsible for the exit poll discrepancy. Of course I do not say "there was no fraud". What is untrue is the assertion that "They never consider that the discrepancies could be due to fraud." They did. I have posted prolifically on this. If you are not lying then the only alternative interpretation is that you haven't bothered to read my analysis. In which case, I suggest you do so.

Febble
They dismiss the pre-election and early exit polls.

"They" certainly do not dismiss the pre-election polls, and if by "early exit polls" TIA means the estimates made on the basis of the polling data alone (and not weighted by the vote returns) this is not true either, as he would know if he had read the E-M evaluation, and indeed, if he'd bothered to read any of my posts.

TIA
Would you care to comment, in detail, on these final national pre-election polls?
http://www.geocities.com/electionmodel/TruthIsAllFAQRes...

Would you care to comment, in detail, on these state pre-election polls?
http://www.geocities.com/electionmodel/TruthIsAllFAQRes...

No. I'd much rather comment (as I have done, extensively) on the actual estimates made by E-M on the basis of the raw exit poll data, as well as the actual raw exit poll data collected at the precinct. As for your pre-election polls, my only comment is that even on your very generous (to Kerry) interpretation of them, the discrepancy between them and the official result is completely uncorrelated with the exit poll discrepancy, suggesting that the discrepancies arise from different causes.

FEBBLE
Not only do "they" NOT disregard this, but they have pointed out that the pre-election polls do not support TIA's case.

Would you care to comment, in detail, on these state pre-election polls?
http://www.geocities.com/electionmodel/TruthIsAllFAQRes...

Done.

FEBBLE
And TIA ignores the experience of world-class pollsters who disagree with him.

TIA
Name them. And explain exactly what they disagree with.

Done already. Ad nauseam. By not only me but many other commentators.

FEBBLE

Again "they" do not. "They" looked very carefully at the historical precedent for incumbents fighting re-lection on a low approval rating, and came to a different conclusion from TIA.

TIA:
"They" don't agree? Exactly who are "they"?

Dunno. Who the hell were you talking about?

This is what Zogby had to say about undecided voters and Bush approval ratings a few days before the election:

http://fpc.state.gov/fpc/37588.htm

And here is what Lou Harris, a world-class pollster with over 40 years experience, said about undecided voters on election day:

http://www.harrisinteractive.com/harris%5Fpoll/index.as...

Are you saying you don't believe these world-class, INDEPENDENT pollsters?
Are you questioning their experience, data, expertise?
If so, tell us why.

No, I am saying that there is more than one reason for a prediction to be in error. And your second link doesn't work

FEBBLE
The truth is that the exit poll evidence does not stack up to evidence of a stolen election - if anything, it contra-indicates the case for theft on a scale of millions of votes. But it does not rule out corruption, and it does not rule out unjust disenfranchisement, particularly of those who had most to gain from a Kerry win. I don't know whether Kerry would have won on a level playing field, but I do know that the playing field wasn't, and isn't, level. And I also know that had it been level in 2000, Gore would be your president now.

But I see absolutely no point in using bad statistical arguments to advance a good cause. TIA's statistical arguments are bad. They don't stand up to scrutiny, and his characterisations of those who have attempted to try to find out what the exit polls actually DID mean are actually dishonest. Good people have spent a lot of time looking at that data. Those people include people who demonstrated conclusively that Gore won Florida in 2000. They do not concur with TIA's conclusions. This is NOT because they were unwilling to consider fraud as a possible source of the exit poll discrepancy. As for me, it was precisely because I was willing to consider fraud as a possible source of the exit poll discrepancy that I ended up analysing the data to try to find out. But the hypothesis was not supported by the data.

TIA
Well, then, scrutinize.
Point out each and every statement and tell us why it is a "bad" argument.
Justify the Lindeman spreadsheet assumptions.

I have been trying to tell you for a couple of years now. I see no point in wasting any more time on it. You show no evidence of actually reading any of my posts anyway.

And aren't you a ghost?