EXIT POLLS: THE LATEST MYSTERY POLLSTER BLOG - AND MY COMMENTS.

[TruthIsAll]

My comments to his points are in bold. The Exit Poll Numbers are what they are. Very compelling circumstantial evidence of fraud. No amount of spinning can make them go away.

The crux of the spin is that we should use a 99.5% condidence interval, and not the standard 95% level. Of course, using 99.5% GUARANTEES THAT EVERY POLL RESULT WILL FALL WITHIN THE MARGIN OF ERROR.

THIS IS TOO, TOO OBVIOUS A CLASSIC DIVERSION TO OBSCURE THESE FACTS: THAT THE BUSH VOTE TALLIES EXCEEDED THE MOE IN AT LEAST SIXTEEN STATES, IF ONE USES THE CALCULATED EXIT POLL MOE IN EACH STATE; OR THAT THEY EXCEEDED THE MOE IN TWENTY-THREE STATES, IF ONE USES THE HISTORICALLY PROVEN 2.0% MOE FOR EXIT POLLS.

http://mysterypollster.typepad.com/

EXITS: WERE THEY REALLY "WRONG?"

Last week's posting of more detailed information on the sampling error of exit polls by the National Election Pool (NEP) allows for a quick review of the now well established conventional wisdom that the "exit polls were wrong."
Conventional wisdom. Of whom?

Let's first set aside the mid-day numbers that were widely leaked on the Internet but never intended as the basis for projections (numbers that even the exit pollsters recognized as flawed in some states -- see this earlier post for explanations on the differences among the many estimates provided by the exit pollsters on Election Day). Let us also stipulate that at least one state estimate of Hispanic voters (Texas) was obviously wrong, given the correction issued by NEP.

The conclusion that the exit polls were wrong is essentially about two possible failings:
• That the end-of-day numbers seemed to predict a Kerry victory.
• That the end-of-day numbers showed Kerry doing consistently better than he actually did.

What is the reality given what we now know about the exit poll's sampling error?

1) Did the just-before-poll-closing exit polls show a consistent and statistically significant "error" in Kerry's favor?

Yes, but that error has been exaggerated. Here is what we know:


• An internal NEP review of 1,400 sample precincts showed Kerry's share of the vote overstated by an average of 1.9 percentage points. As far as I can tell, no one from NEP questions the statistical significance of that overstatement.

TIA:
MY QUESTION TO THE MYSTERY POLLSTER: IS 1.9% OF 1400 SAMPLE PRECINCTS AN INSIGNIFICANT MARGIN?

• The before-poll-closing exit poll results posted by Steven Freeman (and included in the updated report by the Cal Tech / MIT Vote Project) show errors in Kerry's favor in 43 of the 51 surveys (the states plus DC). These overstate Kerry's vote by an average of 1.6 percentage points. If the surveys had been perfectly executed with perfectly random samples (an all but impossible assumption under real world conditions), the pattern of errors should have been the same as if we had flipped a coin 51 times: about half favoring Kerry, about half favoring Bush. The probability of 43 of 51 perfect surveys favoring Kerry by chance alone is less than 0.001%. Of course, this statistic only tells us that the surveys were imperfect. It says nothing about the cause or magnitude of the error.


TIA:
I HAVE CONSIDERED THE MAGNITUDE OF THE ERROR. IT IS BASED ON THE POLLING MARGIN OF ERROR, WHICH THE BUSH VOTE TALLIES EXCEEDED IN 16 STATES AND WHICH THE KERRY TALLIES EXCEEDED IN NONE.


To be clear: Everyone -- including the exit pollsters -- agrees they "overstated" Kerry's vote. There is some argument about the precise degree of certainty of that overstatement, but if all agree that the difference is statistically significant the degree of certainty has little consequence. The size of the error matters, and the reasons for it matter, but whether our level of confidence about the error's existence is 99.9% or something greater does not.

TIA:
THE EXIT POLLS OVERSTATED KERRY’S VOTE?
OR DID THE VOTES UNDERSTATE KERRY’S EXIT POLL NUMBERS?

Having said that, the second draft of the paper, "The Unexplained Exit Poll Discrepancy," continues to needlessly exaggerate the significance of the error, especially within individual states. For example, Freeman claims that there were significant "discrepancies" in Pennsylvania, Ohio and Florida when each state is considered separately. He has a graphic (Figure 1.2) showing a significant error in Florida, assuming a 95% confidence level. However, these assertions are not supported by the margins of error reported by NEP.
• I applied the appropriate "confidence intervals" reported by NEP (as distributed to its partner networks on or before Election Day) to each state. Contrary to Freeman's assertions, the separate "discrepancies" in Ohio, Florida and Pennsylvania fail to attain statistical significance even at a 95% confidence level. In fact, I see only four states (New Hampshire, New York, South Carolina and Vermont) with statistically significant errors favoring Kerry at a 95% confidence level. Of course, when using a 95% confidence level, 2-3 states should be out of range by chance alone.

TIA:
THAT IS FLAT OUT WRONG. ASSUMING THE EXIT POLL MARGIN OF ERROR FOR EACH STATE, KERRY’S VOTE LOSS WAS SIGNIFICANT WHEREVER HIS VOTE PERCENTAGE DEVIATED TO BUSH BEYOND THE MOE – AND THE MOE IS ALWAYS CALCULATED BASED ON THE 95% CONFIDENCE LEVEL, OR +/- 1.96 STANDARD DEVIATIONS FROM THE MEAN.

• I also followed the advice of Nick Panagakis and estimated confidence intervals at a 99.5% level of confidence (the standard used by NEP to make projections) using the actual estimates of the design effect obtained from Warren Mitofsky by blogger Rick Brady. By my calculations, this more demanding test renders the apparent errors in NH, NY, SC and VT non-significant.

TIA:
USE OF THE 99.5% CONFIDENCE LEVEL IS A STRAWMAN ARGUMENT. WHY NOT USE A 99.99% CONFIDENCE LEVEL? THEN YOU CAN BE ASSURED THAT IT WILL NEVER BE EXCEEDED.

Then there is the statistic heard round the world - Freeman's original "250 million to one" estimate of the odds against the discrepancy occurring simultaneously in Pennsylvania, Ohio and Florida. In his second draft, Freeman revised his estimate down to a mere 662,000 to one. Given that Freeman continues to understate the "design effect" used to calculate the sampling error in this year's exit polls, his revised estimate also remains too high.
Some have asked that I calculate my own estimate of the joint probability of an error in Ohio, Florida and Pennsylvania. I am reluctant to do so for two reasons: First, the rounding error in Freeman's data alone renders this sort of hairsplitting moot. Second, and more important, it really doesn't matter. Everyone concedes there was a small (2%) but significant average error in Kerry's direction. For those concerned about problems with the count, what matters most is why that error occurred.

TIA:
DON’T JUST LOOK AT FL AND OH AND PA. LOOK AT ALL 51 STATES. AND CALCULATE THE PROBABILITY THAT 16 OUT OF 51 STATES WOULD MOVE BEYOND THE MOE IN FAVOR OF BUSH.

I AND OTHERS AT DU HAVE CALCULATED THE ODDS THAT THIS COULD OCCUR DUE TO CHANCE:

ONE OUT OF 13.5 TRILLION.

IF YOU CONSIDER THAT FOR EXIT POLLS, THE HISTORICAL NORM MOE = +/- 2%, THEN 23 STATES DEVIATED BEYOND THE 2% MOE IN FAVOR OF BUSH.

THE PROBABILITY OF THAT IS ZERO.


2) Considering sampling error, did the end-of-day numbers predict a Kerry victory?
This one is easy: No. Not even close.

TIA:
OH YES THEY DID.

It is true that the early numbers fooled a lot of people, including the talking heads on the cable networks, pollsters and advisors of both campaigns and possibly -- depending on which accounts you believe -- even John Kerry and George Bush. The reason, as the Post's Richard Morin put it, is that the 1.9% average error in Kerry's favor "was just enough to create an entirely wrong impression about the direction of the race in a number of key states and nationally."

TIA:
ABSOLUTELY UNTRUE. ALL KERRY NEEDED TO WIN WAS EITHER FLORIDA OR OHIO. HE WAS AHEAD IN BOTH. BUSH NEEDED TO WIN BOTH.


True. But that error was not big enough to give Kerry statistically significant leads in enough states to indicate a Kerry victory. If we totally ignore sampling error, the exit polls showed Kerry ahead in only four states that he ultimately lost: Ohio, Iowa, Nevada and New Mexico. Obviously, Kerry would have won the election had he prevailed in all four.

TIA:
AGAIN, NOT TOTALLY ACCURATE. HE JUST NEEDED OHIO.

Before considering whether any of those leads were statistically significant, two quick reminders: First, as Nick Panagakis points out, NEP required at least a 99% confidence level for projections on Election Night. Second, the margin between two candidates on a survey involves two estimates; and the margin of error applies separately to each candidate. Statisticians debate which rules of thumb to apply when determining the significance of the margin between two candidates, but the consensus in the case falls somewhere between the 1.7, as recommended by the American Statistical Association and 2.0, as recommended by those who consider it more appropriate in a race where the vote for 3rd party candidates is negligible (this is a great topic for another day's post - thanks to alert reader Bill Kaminsky for guiding MP through the competing arguments).

TIA:
YOU CONVENIENTLY IGNORE THE DEVIATIONS:
OH 3.0% SIGNIFICANTLY BEYOND THE 2.21% MOE (1963 SAMPLE SIZE)
IA 1.5% CLOSE TO THE 1.97% MOE (2502 SAMPLE SIZE)
NV 2.0% CLOSE TO THE 2.13% MOE (2116 SAMPLE SIZE)
NM 2.0% CLOSE TO THE 2.22% MOE (1951 SAMPLE SIZE)


Fortunately, in this case, the statistical debate is irrelevant. None of Kerry's apparent exit poll leads in the four states were large enough to attain statistical significance, even if we assume a 95% confidence level and use a cautious multiplier (1.7) on the margin of error. As the preceding table shows, the exit polls had Kerry ahead by 4 percentage points in Ohio, by 3 in New Mexico, by 2 in Iowa and by one point in Nevada. The NEP 95% margin of error for these states multiplied by 1.7 is between 5 and 7 percentage points. At the more appropriate confidence level of 99.5% -- the one NEP actually uses to recommend projections - these relatively small margins would fall far short of that needed to call a winner.

TIA:
AGAIN, YOUR USE OF THE 99.5% CONFIDENCE INTERVAL BASED ON THE NEP CRITERIA TO CALL ELECTIONS IS NOT THE ISSUE.

THE ISSUE IS THE CONSISTENT RED SHIFT OF EXIT POLLING RESULTS IN FAVOR OF BUSH – 41 OUT OF 51 STATES, AND 16 ABOVE THE MOE.

******* ONE OUT OF 13.5 TRILLION*******

IS THAT A SIGNIFICANT ENOUGH NUMBER?


In a year-end review in this week's Roll Call (subscription required), Stuart Rothenberg concluded the following:
The problem wasn't the exits - it was the folks who treated the early-afternoon numbers as if they were a predictor of what would happen after everyone had voted. The exit poll was off by a couple of points, but that's well within the margin of error.
If only someone had warned us about those early afternoon numbers on Election Day. Oh, wait...

TIA:
THOSE "EARLY AFTERNOON" NUMBERS (ACTUALLY 4PM) WERE GOOD ENOUGH TO STAY OUT THERE UNTIL MIDNIGHT - AND THEY COMPRISED THE BULK OF THE EXIT POLL SAMPLES.

BUT MITOFSKY WON'T RELEASE THEM. WHY NOT?

UPDATE: I mentioned William Kaminski's blog, but not his commentary on this issue nor his elegant graphic -- it's worth the click just for the chart. My only remaining quibble with his graphic is that it displays a "safe estimate" of a 95% confidence of +/-4%; leaving the impression that the numbers for DE, AL, AK and NE fall outside that interval. However, the appropriate 95% confidence interval provided by NEP for these states is +/-5%.

TIA:
FIRST OF ALL THE CONFIDENCE INTERVAL IS 2.50%, NOT 5.0%.
WE ARE ONLY CONCERNED WITH THE BUSH HALF OF THE NORMAL DISTRIBUTION TAIL. AND IT'S 2.50%, JUST AS IT IS FOR KERRY. THIS IS VERY SOUND, ESTABLISHED AND FUNDAMENTAL STATISTICS. IT'S BASED ON THE EXIT POLL SAMPLE SIZE AND THE RESULTING MOE. THE EXIT POLLS HAD ALREADY ACCOUNTED FOR DEMOGRAPHICS IN THEIR DESIGN AND PREPARATION. AFTER ALL, THAT'S WHAT POLLSTERS ARE PAID FOR, ISN'T IT?

SO, LET ME SAY IT ONE MORE TIME: IN SIXTEEN STATES, THE BUSH VOTING TALLY EXCEEDED THE CALCULATED MOE BASED ON SAMPLE SIZE.
NONE DID FOR KERRY.

THE ODDS: ONE IN 13.5 TRILLION.

AND IN TWENTY-THREE STATES, THE VOTING TALLY EXCEEDED THE HISTORIC, WORLDWIDE, ESTABLISHED 2.0% EXIT POLL MOE FOR BUSH.
JUST TWO STATES FOR KERRY.

THE ODDS: IMPOSSIBLE.

[Goldeneye]

1. The exit polls were used

so that the media could call states for Bush and Kerry. Why would they have bad information on their sites, like the CNN chart that had Kerry winning both men and women in Ohio? It makes no sense. Mitofsky international would have gathered its data, done whatever calculations were necessary, and have sent it to the networks. There is no reason bad data would've made it to CNN. Am I missing anything? It seems like Mitofsky thought their information was fine when they sent it out.

Another question: did Mitofsky come out and say they thought their exit polling was biased?

[BeFree]

6. Bad data?

Yeah, right. The news groups paid $10 million dollars for what they figured was gonna be bad data? I don't think so. I think they figured the data they paid 10 mil for was gonna be right on, or at most within 2 points.

And now they want us to believe they spent 10 mil for data that they thought would be wrong as much as 3, 4, 5, even 8 percent? I don't think so.

They aren't idiots, but they do play one on TV, eh?

[bemis12]

9. It's 3% or more

according to the guy doing the polls.

http://www.exit-poll.net/faq.html#a15

[jkd]

23. 4%

Mitofsky says the MOE for a state poll is + or - 4%. If there was an additional built-in bias of 2% for Kerry, what does that tell us about the reliability of these exit polls? The bias was actually higher than the 2% average in some states.

[TruthIsAll]

24. No! 4% is MOE for a standard poll-600 sample size. Exit polls are better

There is no end to the misrepresentations and spin.
But ask any pollster. He'll tell you it's BS.

Mitofsky exit-polled 2816 in FL. That sample size gives a 1.84% MOE, based on standard pre-election telephone polling. Exit polls are more accurate than standard polls.

Want to calculate the MOE for a standard poll of any sample size (N)? It's easy.

Say N= number polled (sample size).
The MOE = 1/sqrt(N).
It's that simple.

For FL, the MOE = 1/Sqrt(2816) = .0184.
Tight fit. Feels good.
It means that 95% of the time, the sample mean will be within 1.84% (+ or -)of the population mean.

Mitofsky mispoke. He is spinning. In fact, he always claimed a 3% MOE, but he has to widen it now and claim it's bad sampling.

Right.

[jkd]

27. You can believe that, if you wish.

Mitofsky interviewed far fewer than 2800 in most of the states. In my state of Utah it was about 700. The BYU exit polls were within .5% and Mitofsky missed it by 4.5%. Did Mitofsky just screw-up or is he lying?

[TruthIsAll]

30. Utah? Talk FL, OH. They were the ones Bush needed. n/t

.

[jkd]

31. Did he only try in Florida and Ohio?

If he did a poor job in Utah, can we believe that he did any better in the other states?

[berniew1]

71. Vote machine fraud has been documented in swing states-explains swing

Systematic vote machine fraud, "Default to Bush" has been documented in swing states such as Ohio, Florida, and New Mexico- and some of this has been observed in other states as well.
http://www.flcv.com/fraudpat.html
Other types of systematic fraud have also been observed in many areas.
These appear to explain the swing to Republican seen by the difference in the exit polls and official vote totals.

This means that Bush or another default candidate would get not only the votes of those who didn't notice that the intended candidate didn't register, but any intentional non votes or accidental non votes or unsuccessfully completed votes. Quite an advantage. Under normal circumstances this would result in a swing as much as 2 to 4% in many areas. But since it is a matter of record that the default was hard to override and sometimes impossible as reported by the EIRS reports and acknowleged by poll workers, only the most watchful or purposeful voters may have been able to successfully vote for a chosen non-Republican candidate. Thus the swing was likely much larger for some machines and even more in some minority precincts that appear to have been targeted to produce high levels of misvotes based on the large number of EIRS reports in minority precincts. There were also a few cases where precincts had machines set to default for a minor party candidate, but none observed to default to Kerry. These patterns have been confirmed by analysts and computer experts in several counties of each of these states. Other similar default related patterns have also been observed in some of these areas, such as failure to register a presidential vote when a voter tried to vote straight party line. This would result in the presidential vote going to the other “unintended” party.

[jkd]

75. No defaults

Utah uses punch-card ballots. There are no defaults. This doesn't explain Mitofsky missing in Utah by 4.5%

[jkd]

78. Re-poll Ohio

davidgmills proposal to re-poll Ohio may have some merit, if Mitofsky will not release his data. The interviews must be completely at random among those who actually voted. One would need the poll books to obtain this information. Will the administrators allow public access to all of these books?

The number of interviews would have to be greatly increased because post-election polls are much less accurate than well-done exit polls. If Mr. Blumenthal says that 16,000 is enough, he would know.

[Pierre de Fermat]

25. The media paid 10 million for...

....a exit poll and they got one. The 'raw data' you mention are the exit poll raw numbers. Raw exit poll numbers are one thing, and the exit poll is another. Many people conflate the two which is incorrect. While the media did pay for the raw numbers, they hold very little value for them. The media is after the actual exit poll, which enables them to say things like "In the midwest, hispanic voters voted 2 to 1 for Kerry."

You may ask, how can the exit poll be of any use if the raw numbers from which it is derived are not accurate ? There's an answer - use google.

The bottom line is simply this: The exit poll raw numbers can't be used the way some here on DU are attempting to use them. because of this, most of the 'analysis' done by DU posters seems to be void of any value. The media understands that the raw numbers are meaningless because they've been paying for exit polls for years. This helps to explain why they aren't covering the fraud angle.

I've read the analysis on DU and I have read Mystery Pollster. Sorry, TIA, MM outranks you.

[TruthIsAll]

29. Your opinion is void of the truth. Fermat was a great mathematician.

No one has refuted the analysis.

A poll is a poll. Kerry was winning. Since when do we have to wait for exit poll results until after midnight?

Get real.

Don't try to pull rank on me. I have been doing mathematical modeling and programming for probably more years than you have been alive.

And no one is paying me for this, either.

[On the Road]

36. Your Analysis

gives only the mathematical part of the picture. The margin of error applies only if what you're sampling is the actual data, which in this case would be the actual ballots.

It may not enter into purely mathematical modeling, but social scientists like pollsters and psychometricians know this. Once you stop people and ask their responses, you get a whole new set of sampling and response variables.

I don't know how accurate the exit polls were. The recount in New Hampshire examined wards in which the official vote deviated quite a bit from the polls. Presumably those were not the only places there was a difference.

A better way of isolating fraud might be to look at precincts which had especially high deviations. Which would only be possible if Mitofsky released the details.

[gulogulo]

37. "The recount in New Hampshire examined wards"

" in which the official vote deviated quite a bit from the polls"

That is not the impression I had of the NH recount. AFAIR Nader stated that there were no significant differences found in the recount. Do you have any link to show that there were significant deviations found?

[On the Road]

44. Here's the Kind of Article I Was Referring to

Recount New Hampshire
by Russ Baker

----snip

Tomorrow the first recount begins--in New Hampshire, of all places, a state George Bush didn't even win. But in those areas where he did well, sometimes the numbers look decidedly odd. In this case, the person who got the ball rolling was one Ida Briggs, a longtime Michigan software designer and database developer who did a statistical analysis of some election results, and found them perplexing enough to trigger concerns in her mind about the efficacy of the electronic vote tabulation system used.

----snip

Briggs became interested in the numbers when, shortly after the election, she saw a study published on the web about statistical anomalies in nonswing states. New Hampshire caught her attention because of the sizable--15 percent--differential between early exit polls and results. It was easy to study, because the state made its data available online. And because New Hampshire was a state Kerry won, no one could claim that the goal of a recount there was to change the election results

http://www.commondreams.org/headlines04/1118-03.htm

The point is that on a limited basis, the exit poll discrepancy was checked and resolved in favor of the official results. Doesn't mean it would match everywhere, but it does mean that it can happen, and that sampling votes via exit poll is not the same as sampling the actual ballots.

[Duncan]

42. If raw numbers were released

If raw numbers were released it would be along with demographic info that pollsters and statisticians like TIA have accounted for and would continue to account for, and I suspect TIA's analysis would likely be strengthened by such a release.
WHY WON'T POLLSTERS RELEASE THEIR DATA NOW? Nobody can say anything about how "raw" any data is until the pollsters release their data.
How "raw" was the data that led Powell and others, Republican and Democrat, to declare fraud occurred in Ukraine?

The shear magnitude and expanse of the Bush shift from exit polls makes it blatantly clear that fraud occurred, even if you aren't a statistician. You don't even have to play poker to see that we were cheated. The disgust with this fact is not going to go away. It is going to grow. Nobody can spin their way out of the numbers, and the numbers are not going away.

[Election Mess]

56. MSM should demand refund from Mitofsky for shoddy exit poll work

MSM should demand refund from Mitofsky. Maybe then he'll start to defend his data. That would blow the doors off this thing.

[Nederland]

10. Yes

Another question: did Mitofsky come out and say they thought their exit polling was biased?

Yes, Mitofsky explains why the exit polls were flawed here (his comments are mostly on page 2):

http://www.washingtonpost.com/wp-dyn/articles/A64906-2004Nov20.html

[BeFree]

12. Flawed?

He oughtta give 'em back their money then.

If he screwed up, and he admits he did, he should do the honorable thing by giving back the money and releasing the raw data.

Wouldn't you like the real facts?

[Nederland]

38. Agreed

Mitofsky should give the money back. He produced a lousy product and doesn't deserve to be paid.

[impeachthescoundrel]

2. Isn' t exit polling conducted

after the vote is cast? Even if you had been under a rock. you would know better than that. This election was stolen just like the last one; of course that is only my opinion, but I am not the only one of this opinion!

[berniew1]

3. Exit polls have been documented correct in 2000 and 2004

Media recounts after 2000 election documented that Gore would have won Florida by a lot in a fair count; and likewise the polls have been documented to be simiarly correct in 2004

Widespread patterns of vote machine fraud have been documented in many states in 2004 www.flcv.com/fraudpat.html
and
There was also widespread systematic voter suppression of minority voters in many states which was well documented from the EIRS cases
https://voteprotect.org
as well as the following cases and analysis, which also have further documentation on the vote machine manipulation patterns.

Ohio
www.votersunite.org/electionproblems.asp?sort=date&selectstate=OH&selectproblemtype=ALL
http://northnet.org/minstrel/columbus.htm
http://northnet.org/minstrel/cleveland.htm
http://www.flcv.com/ohiov04.html
New Mexico
http://www.helpamericarecount.org/NewMexicoData/NewMexicoGeneralElection.pdf
http://www.votersunite.org/electionproblems.asp?sort=date&selectstate=NM&selectproblemtype=ALL
Florida
http://www.votersunite.org/electionproblems.asp?sort=date&selectstate=FL&selectproblemtype=ALL
http://www.flcv.com/fla04EAS.html
http://www.flcv.com/flavi04.html

[TruthIsAll]

4. MUST READ for DUers to understand the BushCo Exit Poll Spin Machine

We have got them dead to rights.

[RubyCat]

5. kick

[procinderella]

7. Read the petition -- it lays out a similar case.

READ the Petition filed with the Ohio Supreme Court. After you get to about page 15 or sixteen, it starts looking at statistical analysis of the exit polling data and the "official" returns.

It also talks about the small window that the "real" exit polls were posted. These exit poll numbers were later changed and called "exit poll" numbers when combined with "official" numbers.

www.freepress.org

I hypothesize the the actual result of the election should have been Kerry at 61,000,000 and Bush at 54,000,000.

[Vinnie From Indy]

8. Thank You TIA

Your posts are always informative and eagerly awaited!

[fshrink]

11. Although all this is true, I still think

that the best strategy is to look for patterns, which statictics does allow. At the source of the fraud there seems to have been a directive to aim at 5% error max. Bif enough to make a difference, small enough to slip by. The second directive was probably to use several different methods so that the deviation couldn't be easily identified. However, among all the means available the electronic systems offered the best reliability and the best fine-tuning. Slowing down turn-out, for example, although efficient, is risky and not quantifiable. Therefore, there must be a correlation between the amount of "error" and the presence of electronic devices. This can be proven with different methods, the simplest of which probably is a standard Chi-square. This is not sophisticated probabilistic analysis, just a run of the mill, but sensitive enough, standard statistical method. And if there has been fraud, I bet a Chi-Square will give you a significant result at the .01 level.

[Heewack]

13. mystery pollster says:

{Exit poll} reliability can be questionable. One might think that there is no reason why voters in stable democracies should conceal or lie about how they have voted, especially because nobody is under any obligation to answer in an exit poll. But in practice they often do. The majority of exit polls carried out in European countries over the past years have been failures {emphasis added}.

[TruthIsAll]

15. Is that so? They could have fooled me. Mystery Pollster, source it.

That is one helluva statement.
Remarkable, really.

I guess we might as well close up shop now.

Exit polls are consistently wrong. Sheesh.

The Germans have been lying to us. And so have the Ukrainians. And the French. And the Georgians.

Right.

[Heewack]

16. I'm just copying from the link you posted

I have no reason to doubt his statement. He says that exit polls over the past few years in Europe have been failures. That seems to be the case here as well at least to some extent. I think that is worthy of some consideration. Do you not?

[gulogulo]

26. I think Mystery Pollster does source it

follow the links he cites on:

http://www.mysterypollster.com/main/2004/12/what_about_thos.html

he cites this, for example:

http://www.aceproject.org/main/english/lf/lfd08e.htm

"...their reliability can be questionable. One might think that there is no reason why voters in stable democracies should conceal or lie about how they have voted, especially because nobody is under any obligation to answer in an exit poll. But in practice they often do. The majority of exit polls carried out in European countries over the past years have been failures."

(ACE Project is a joint venture of the International Foundation for Election Systems (IFES), the International Institute for Democracy and Electoral Assistance (IDEA) and the United Nations Department of Economic and Social Affairs (UNDESA))

he also quotes from the email of Dr. Dieter Roth of FG Wahlen, who conduct German exit polls, who says:

"I know that Warren Mitofsky's job is much harder than ours, because of the electoral system and the more complicated structure in the states."

[Pierre de Fermat]

28. This is a well known phenomenon. It's simple to find studies...

...online about non-response and dishonesty in exit polling. Mystery Pollster is correct. Not only is it corrolated to conservatives, it's also correlated to age and time of day. For example:

http://www.cpsa-acsp.ca/papers-2004/Docherty%20and%20others.pdf

[Nederland]

39. Thanks

That was a fascinating paper. It certainly proves that exit polls are not as accurate as some here have claimed...

[breadbox]

14. How does MysteryPollster respond?

TIA,
Have you discussed your differences with MysteryPollster? He seems
to be pretty open to debate, and has a relatively open mind --- I think
that he has probably concluded that there was no fraud on the levels
that some believe, but that if you have strong evidence you could
convince him. He is a good guy, and on the right side: a good debate
on your observations could be very useful: it could strengthen your
arguments (and at the same time, if there are weaknesses, expose them,
always useful ahead of time).

Breadbox

[TruthIsAll]

17. To his credit, he has posted my numbers in his blog.

But he does not appear to do his own analysis; rather, he gets his info elsewhere. Nothing wrong with that.

But I question his level of mathematical knowledge.

That 99.5% confidence level is just a confidence con. He apparently fell for it.

The probabilistic analysis of exit poll deviations has NOT A DAMN THING TO DO WITH CALLING THE ELECTIONS ON NATIONAL TELEVISION.

He is spinning very hard to make it seem that the deviations were no big deal. They were. Very big. In both number AND magnitude.

Just use the binomial distribution:
Sixteen (16) out of 51 independent events occurred, each with a probability of .025.

The odds of AT LEAST 16 out of 51:
****** ONE OF 13.5 TRILLION ******

Use your favorite spreadsheet (I use Excel).
Prob = 1-BINOMMDIST(15,51,.025,true)

Odds = 1/Prob

DO THE MATH.
I DID.

[Heewack]

18. But which poll numbers are you using?

The 4 p.m. numbers?

[TruthIsAll]

19. Exactly. Are there any others? Ask Mitofsky? Conyers already did

Oh, I forgot about the contaminated CNN numbers.
Sorry. Can't use 'em.

I like 'em clean. Untouched by sub-human hands.

tia

[Nederland]

45. The 4pm Numbers

You know, the ones that everybody except TIA has already admitted were inaccurate. Oh wait, I forgot, Dick Morris liked them too.

[TruthIsAll]

48. Those 4pm numbers represent 71,492 individuals polled nationwide

Nice sample (oh, response) size. No undecided voters here.

They voted, Dems AND Repubs, and others.

And there is no reason to believe they were not being truthfull.

[gulogulo]

49. there is also no reason to believe

that they were representative.

[Nederland]

59. Hasn't anyone ever told you

size isn't everything? :)

[breadbox]

46. doing the math

TIA,
I'm not good at writing gently: I write from the heart, and if any
of the below appears to be criticizing you, please please please
don't take it that way --- it is not meant to. I have a tremendous
respect for your postings over the past few weeks, and for your
investigations. I really would like to see you and MP get a dialog
going that could lead to a brighter future than to focus on the
disagreements in public.

About me: I am a professional mathematician: not a statistician,
but in a related field: for reasons of anonymity I will say no more
(already the nickname and profession would identify me to a good
number of people I know!)

I've done rather a lot of the math recently (in particular, teaching
two classes about exit polls under perfect circumstances --- and at
the moment attempting to do a combinatorial treatment of the choosing
precincts in a non-ideal world where the choice of precincts can affect the outcome).

I read through MP's post fairly carefully since it is the exit polls
that brought me to the conclusion on Nov 3 that the election had been
rigged --- after Avi Rubin's articles had demonstrated how easy some
bits of the rigging would be to do. I don't like being a tin foil hatter, so I am trying to weigh evidence on both sides: much as I wish
Kerry had won, if someone were to *prove* to me that the election was
fair, it would please me for the sake of democracy. I'd cry for our
country and its choices, but I would breathe a breath for the sake of
democracy.


I think that you and MP are arguing different numbers at each other:
I understand your odds of 16 out of 51 rare events happening, and
think that it is reasonable to assume independence of the variables
--- I would modify the computation just slightly, to compute the
odds of *at least* 16 events happening, which will increase the odds
slightly, but they will still be drastically small.

I think that what MP has computed is something different: he is
assuming that the odds of a state exit poll showing a bias for X over
Y is 1/2: then the odds that (again, we should use at least rather
than exactly, which changes the odds, but....) 43 out of 51 polls
would all show a bias in favor of X is less than 0.001%.
I haven't bothered to redo this calculation: it feels reasonable.

These two things are completely different statistics. The nice thing
about *your* statistics is that the odds that you get are extremely
small that this could happen by chance. However, to perform the
calculation properly, you do need to know far more about MOEs than
I (or NEP or anyone else, for that matter) know when you are dealing
with precinct selection, etc. I think that MP has done a good job of
explaining why some of the MOE calculations are not as trivial as
those I do in class (I agree in principle --- I don't yet know if I
agree with the numbers).
The nice thing about *his* statistics is that, a priori, there is no
reason to assume one candidate or the other should be ahead in exit
polls, assuming that the precincts are numerous and randomly selected,
etc. Hence the odds that 43 out of 51 should fall on one side of
a line can be computed, and it really is a very meaningful number.
It is not as small as yours, but it is small enough that we wouldn't
expect to see it happen by chance in 100 years of elections. (even
in a millenium of elections, but it is beginning to get close).

It is adequately small for us to scream blue murder, in my opinion.

I think that you are correct in that MP is not a mathematician -- he
is a professional pollster, used to interpreting common events, not
rare events. My suggestion to you is that you communicate gently with
him, leading him to see that your ideas are orthogonal to what he has
discussed, and are worthy of discussion.

He is well regarded in the mainstream left, those who are ignoring
the fraud are doing it at least partly because he isn't telling them
that it looks real: persuade him and we have a good ally.

All the best, and keep up the good work,
Breadbox

[TruthIsAll]

51. I appreciate your comments.

The probability calculation IS for AT LEAST 16 state vote tallies to exceed the polling MOE - all favor of Bush.

I used the the Excel Binomial Distribution probability function.

This function returns the individual term binomial distribution probability. Use BINOMDIST in problems with a fixed number of tests or trials, when the outcomes of any trial are only success or failure, when trials are independent, and when the probability of success is constant throughout the experiment. For example, BINOMDIST can calculate the probability that two of the next three babies born are male.

Syntax
BINOMDIST(number_s,trials,probability_s,cumulative)

Number_s is the number of successes in trials.
Trials is the number of independent trials.
Probability_s is the probability of success on each trial.

Cumulative is a logical value that determines the form of the function. If cumulative is TRUE, then BINOMDIST returns the cumulative distribution function, which is the probability that there are AT MOST number_s successes; if FALSE, it returns the probability mass function, which is the probability that there are number_s successes.

Let X = The probability that at MOST 15 would fall outside the MOE:
X = BINOMDIST(15, 51, .025, TRUE).

Here .025 represnts the probability that a state would deviate beyond the mOE in favor of Bush.

So the converse probability P that AT LEAST 16 would fall outside the MOE is: P = 1 - X

The odds are equal to 1/P or 1 out of 13.5 trillion.

*********************************************
The odds that at least 41 out of 51 states would move in favor of Bush (without regard to magnitude) from the exit polls is 1 out of 135,000.

Once again, we use the binomial distribution, with .50 as the probability that a given state would move to Bush from the exit polls.

Let X = The probability that at MOST 40 would fall outside the MOE:
X = BINOMDIST(40, 51, .025, TRUE)

So the probability P that AT LEAST 41 would fall outside the MOE is:
P = 1 - X

The Odds = 1/P = 1 out of 135,000

As far as the analysis is concerned:

1. The MOE for a standard (non-exit) poll is strictly a function of the number sampled. The MOE = 1 /Sqrt(N)
For 2826 sampled in FL, the MOE is 1.84%

2. Historic experience shows that Exit Polls are more accurate than standard polls.

3. Over 71,000 individuals were exit-polled nationwide in the exit poll data downloaded by Simon at 12:22 AM on Nov 3.

I used these nnumbers to calculate the MOE and vote tally deviation from the exit poll for each state, as I have displayed above.

Any attempt at spinning the Kerry lead in the exit polls as being due to early voting; incomplete, biased samples; Bush voters coming late to the polls; Kerry's voting early; Bush voters not talking to pollsters, Kerry's doing so.....etc, etc. is designed to not get at the truth, but to rewrite history.

It's truly Orwellian.

I have been a quantitative analyst/model builder all my life, and have worked in many disciplines. I do not pretend to be a polling expert, but I know that the basic analysis is correct. This is NOT rocket science. I have several advanced degrees in mathematics, and I know from long-time HANDS-ON experience that my analysis is essentially correct.

[breadbox]

52. not an excel user

TIA: I'm not an excel user so I missed the cumulative flag
(although I do recall now seeing that you mentioned it in an
earlier post either in this or an earlier thread).

On problem with MOE for exit polls is the precinct selection:
to take an extreme example, consider a state with two precincts,
one all R, the other all B: now if we sample just one precinct the
result will be all R or all B. If we sample two precincts, then
the result will closely mirror the relative sizes of the precincts,
right? Well, it will, so long as the sample sizes in the precincts
match the voting population sizes of the precincts.

When it is as easy to construct pathological examples such as these,
there tend to be subtleties around which can kick you in the rear.

I really am not trying to spin anything here: I tend to go with
Occam's razor for problems: if there is a problem, try the easiest
explanation first: it will often be the truth. Unfortunately,
(I say this because it sickens me that I think that they did this to
our country) unfortunately, the simplest explanation seems to me to
be massive, widespread, very narrowly focussed, multi-headed fraud.
And when Occam comes up with a sentence like that last one, you can
be sure that others are going to question its simplicity.

Keep up the good work, TIA.
Breadbox

[TruthIsAll]

57. If you tend to go with Occam's razor, then you must believe Bush stole it.

That is the most likely explanation, isn't it?

I base this on election 2000 and 2002, both stolen. Why should 2004 be any different? That is Occam.

I base it on the fact that he has cheated on everything from his AWOL days right up to the minute. That is Occam.

I base it on the literally thousands of lies from BushCo. Why should he not lie about this, too? That is Occam.

I base it on the simple fact that his ratings were and still are below 50%. That is Occam.

I base it on the simple fact that as an incumbent who was essentially tied at 47 in the polls, he would lose the undecided vote. That is Occam.

I base it on the simple fact that Exit Polls are used throughout the world because they are accurate. That is Occam.

I base it on the simple fact that Mitofsky won't release his numbers or appear at the hearing when asked by Conyers. That is Occam.

And Bush will die from a thousand cuts of Occam's razor.

[TruthIsAll]

58. TYPO FIX: X = BINOMDIST(40, 51, .5, TRUE)

Let X = The probability that at MOST 40 would fall outside the MOE
should be:

X = BINOMDIST(40, 51, .5, TRUE)

Here 0.5 is the probability. Flip a coin.
I inadvertantly copied .025 in the function by mistake.

The odds are correct: 1 in 135,000

[Ellipsis]

20. Gotta send this one out, ...speechless at the talent at DU! n/t

[shiina]

21. Link?

Have you gone into detail about how you got that number (13.5 trillion) somewhere? Could you put a link to it? I would be really interested in seeing that.

Thanks.

[TruthIsAll]

22. Here it is.

Just use the binomial distribution: 
Sixteen (16) out of 51 independent events occurred, each with
a probability of .025.

The odds of AT LEAST 16 out of 51:
****** ONE OF 13.5 TRILLION ******

Use your favorite spreadsheet (I use Excel).
Prob = 1-BINOMDIST(15,51,.025,true)

Odds = 1/Prob


Here are the Exit Poll and Voting Results for all the states:

Size refers to the exit poll sample size for the given state.
The percentages are Kerry's Exit Polls and reported Votes.

State	Size	Exit	Vote	Diff	StDev	MoE	Prob  	>MoE?	Favor
DE	770	58.50%	53.54%	-4.96%	1.80%	3.53%	0.29	yes	Bush
NH	1849	55.40%	50.51%	-4.89%	1.16%	2.28%	0.00	yes	Bush
VT	685	65.00%	60.20%	-4.80%	1.91%	3.74%	0.60	yes	Bush
SC	1735	46.00%	41.41%	-4.59%	1.20%	2.35%	0.01	yes	Bush
NE	785	36.76%	32.32%	-4.44%	1.78%	3.50%	0.64	yes	Bush

AK	910	40.50%	36.08%	-4.42%	1.66%	3.25%	0.38	yes	Bush
AL	730	41.00%	37.00%	-4.00%	1.85%	3.63%	1.53	yes	Bush
NC	2167	48.00%	44.00%	-4.00%	1.07%	2.11%	0.01	yes	Bush
NY	1452	63.00%	59.18%	-3.82%	1.31%	2.57%	0.18	yes	Bush
CT	872	58.50%	55.10%	-3.40%	1.69%	3.32%	2.24	yes	Bush

RI	809	64.00%	60.61%	-3.39%	1.76%	3.45%	2.68		Bush
MA	889	66.00%	62.63%	-3.37%	1.68%	3.29%	2.21	yes	Bush
PA	1930	54.35%	51.00%	-3.35%	1.14%	2.23%	0.16	yes	Bush
MS	798	43.26%	40.00%	-3.26%	1.77%	3.47%	3.29		Bush
OH	1963	52.10%	49.00%	-3.10%	1.13%	2.21%	0.30	yes	Bush

FL	2846	50.51%	47.47%	-3.03%	0.94%	1.84%	0.06	yes	Bush
MN	2178	54.50%	51.52%	-2.98%	1.07%	2.10%	0.27	yes	Bush
UT	798	30.50%	27.55%	-2.95%	1.77%	3.47%	4.78		Bush
ID	559	33.50%	30.61%	-2.89%	2.11%	4.14%	8.60		Bush
AZ	1859	47.00%	44.44%	-2.56%	1.16%	2.27%	1.38	yes	Bush

VA	1000	47.96%	45.45%	-2.50%	1.58%	3.10%	5.66		Bush
LA	1669	44.50%	42.42%	-2.08%	1.22%	2.40%	4.49		Bush
IL	1392	57.00%	55.00%	-2.00%	1.34%	2.63%	6.78		Bush
WI	2223	52.50%	50.51%	-1.99%	1.06%	2.08%	3.00		Bush
WV	1722	45.25%	43.43%	-1.82%	1.20%	2.36%	6.54		Bush

NM	1951	51.30%	49.49%	-1.81%	1.13%	2.22%	5.54		Bush
CO	2515	49.10%	47.47%	-1.63%	1.00%	1.95%	5.15		Bush
IN	926	41.00%	39.39%	-1.61%	1.64%	3.22%	16.42		Bush
GA	1536	43.00%	41.41%	-1.59%	1.28%	2.50%	10.69		Bush
MO	2158	47.50%	46.00%	-1.50%	1.08%	2.11%	8.17		Bush

NJ	1520	55.00%	53.54%	-1.46%	1.28%	2.51%	12.67		Bush
WA	2123	54.95%	53.54%	-1.41%	1.09%	2.13%	9.70		Bush
IA	2502	50.65%	49.49%	-1.15%	1.00%	1.96%	12.41		Bush
AR	1402	46.60%	45.45%	-1.15%	1.34%	2.62%	19.55		Bush
KY	1034	41.00%	40.00%	-1.00%	1.55%	3.05%	26.01		Bush

OK	1539	35.00%	34.00%	-1.00%	1.27%	2.50%	21.63		Bush
MI	2452	52.50%	51.52%	-0.98%	1.01%	1.98%	16.47		Bush
NV	2116	49.35%	48.48%	-0.87%	1.09%	2.13%	21.29		Bush
ME	1968	54.75%	54.08%	-0.66%	1.13%	2.21%	27.80		Bush
MD	1000	57.00%	56.57%	-0.43%	1.58%	3.10%	39.18		Bush

DC	795	91.00%	90.91%	-0.09%	1.77%	3.48%	47.96		Bush
MT	640	39.76%	39.80%	0.04%	1.98%	3.87%	50.72		Kerry
OR	1064	51.20%	52.00%	0.80%	1.53%	3.00%	69.91		Kerry
HI	499	53.30%	54.55%	1.25%	2.24%	4.39%	71.10		Kerry
TX	1671	37.00%	38.38%	1.38%	1.22%	2.40%	87.10		Kerry

TN	1774	41.50%	43.00%	1.50%	1.19%	2.33%	89.68		Kerry
CA	1919	54.00%	55.56%	1.56%	1.14%	2.24%	91.35		Kerry
SD	1495	37.76%	39.39%	1.63%	1.29%	2.53%	89.65		Kerry
ND	649	34.00%	36.36%	2.36%	1.96%	3.85%	88.58		Kerry
KS	654	35.00%	37.37%	2.37%	1.96%	3.83%	88.76		Kerry

Avg	1450	49.18%	47.38%	-1.80%	1.42%	2.79%	21.67		Bush
Med	1507.5	49.23%	47.47%	-1.81%	1.29%	2.52%	6.66		Bush
									
						

[platinumman]

32. Please justify your assumptions

Every time you've posted this stuff in the past week or so, I have asked you to defend your assumption that precincts were randomly sampled. So far you have failed to do so.

If you don't know the sampling methods, you cannot make any calculations at all regarding the margin of error. The only way you can possibly know the design of the exit polling is if you had some inside knowledge. Do you?

Just for the factual record, commercial polling companies would rarely sample completely randomly. Of course, you may be right in this case, but you cannot claim to be right unless you give your sources.

If you keep posting this stuff, I'll just keep asking you to justify your assumptions. If you can't, your calculations should be dismissed.

[TruthIsAll]

33. Please do your homework.

I have not seen your posts asking for justification of the
random samples. I do not avoid questions. But for you to ask
me at this point what is my source, I can only assume you are
unfamiliar with my posts. The info has been out there for a
month. The source is Simon/SCOOP. No one has said the numbers
were incorrect.

I am not a polling expert. But I know enough math to know when
I am being fed pablum. The timing of the exit polling  was no
different than in the past. Why was the media quoting them if
they were bogus? 

The networks have always have made their election winner calls
based on 99.5% confidence. But as you are well aware, the
99.5% coinfidence level is not applicable to polling MOE. Here
a 95% confidence level used.

As for my MOE assumptions, I base it on the sample size as
presented by Simon and SCOOP:

http://www.democraticunderground.com/discuss/duboard.php?
az=view_all&address=203x36314

And are you saying that the exit poll is any LESS accurate
than standard polls, which typically have a 2.5-3% MOE?

Tell me why I should not use the numbers I did?
Tell me why Mitofsky won't release the data?
Tell me why you question the probability calculation? Isn't
that what it is all about? 

Your argument is a combination of: 
1) The Polling data used was insufficient.
2) The Polling methodology was flawed.
3) The MOE was over 4%.

Which is it? You can't have it three ways. 

I will NOT let you get away with the 4%+ MOE. That is a much
too high an MOE for Exit Polls. It's applicable to standard
state polls with a 600 sample size.

Exit polls have been extremely accurate historically - less
than 2% MOE. 

The MOE for standard polls =1 /Sqrt (N) where N= sample size.
That is how I compute the MOE. The sample size for FL was 2816
and 1963 for OH.

Check actual historical polls (Gallup, Zogby - all of them).
Apply the MOE formula using the published sample size. You
will find it ALWAYS MATCHES THE PUBLISHED MOE.

WHY SUCH A HULLABALOO ABOUT THE CALCULATED MOE?

Here are the Exit Poll and Voting Results for all the states:

Size refers to the exit poll sample size for the given state.
The percentages are Kerry's Exit Polls and reported Votes.

State	Size	Exit	Vote	Diff	StDev	MoE	Prob  	>MoE?	Favor
DE	770	58.50%	53.54%	-4.96%	1.80%	3.53%	0.29	yes	Bush
NH	1849	55.40%	50.51%	-4.89%	1.16%	2.28%	0.00	yes	Bush
VT	685	65.00%	60.20%	-4.80%	1.91%	3.74%	0.60	yes	Bush
SC	1735	46.00%	41.41%	-4.59%	1.20%	2.35%	0.01	yes	Bush
NE	785	36.76%	32.32%	-4.44%	1.78%	3.50%	0.64	yes	Bush

AK	910	40.50%	36.08%	-4.42%	1.66%	3.25%	0.38	yes	Bush
AL	730	41.00%	37.00%	-4.00%	1.85%	3.63%	1.53	yes	Bush
NC	2167	48.00%	44.00%	-4.00%	1.07%	2.11%	0.01	yes	Bush
NY	1452	63.00%	59.18%	-3.82%	1.31%	2.57%	0.18	yes	Bush
CT	872	58.50%	55.10%	-3.40%	1.69%	3.32%	2.24	yes	Bush

RI	809	64.00%	60.61%	-3.39%	1.76%	3.45%	2.68		Bush
MA	889	66.00%	62.63%	-3.37%	1.68%	3.29%	2.21	yes	Bush
PA	1930	54.35%	51.00%	-3.35%	1.14%	2.23%	0.16	yes	Bush
MS	798	43.26%	40.00%	-3.26%	1.77%	3.47%	3.29		Bush
OH	1963	52.10%	49.00%	-3.10%	1.13%	2.21%	0.30	yes	Bush

FL	2846	50.51%	47.47%	-3.03%	0.94%	1.84%	0.06	yes	Bush
MN	2178	54.50%	51.52%	-2.98%	1.07%	2.10%	0.27	yes	Bush
UT	798	30.50%	27.55%	-2.95%	1.77%	3.47%	4.78		Bush
ID	559	33.50%	30.61%	-2.89%	2.11%	4.14%	8.60		Bush
AZ	1859	47.00%	44.44%	-2.56%	1.16%	2.27%	1.38	yes	Bush

VA	1000	47.96%	45.45%	-2.50%	1.58%	3.10%	5.66		Bush
LA	1669	44.50%	42.42%	-2.08%	1.22%	2.40%	4.49		Bush
IL	1392	57.00%	55.00%	-2.00%	1.34%	2.63%	6.78		Bush
WI	2223	52.50%	50.51%	-1.99%	1.06%	2.08%	3.00		Bush
WV	1722	45.25%	43.43%	-1.82%	1.20%	2.36%	6.54		Bush

NM	1951	51.30%	49.49%	-1.81%	1.13%	2.22%	5.54		Bush
CO	2515	49.10%	47.47%	-1.63%	1.00%	1.95%	5.15		Bush
IN	926	41.00%	39.39%	-1.61%	1.64%	3.22%	16.42		Bush
GA	1536	43.00%	41.41%	-1.59%	1.28%	2.50%	10.69		Bush
MO	2158	47.50%	46.00%	-1.50%	1.08%	2.11%	8.17		Bush

NJ	1520	55.00%	53.54%	-1.46%	1.28%	2.51%	12.67		Bush
WA	2123	54.95%	53.54%	-1.41%	1.09%	2.13%	9.70		Bush
IA	2502	50.65%	49.49%	-1.15%	1.00%	1.96%	12.41		Bush
AR	1402	46.60%	45.45%	-1.15%	1.34%	2.62%	19.55		Bush
KY	1034	41.00%	40.00%	-1.00%	1.55%	3.05%	26.01		Bush

OK	1539	35.00%	34.00%	-1.00%	1.27%	2.50%	21.63		Bush
MI	2452	52.50%	51.52%	-0.98%	1.01%	1.98%	16.47		Bush
NV	2116	49.35%	48.48%	-0.87%	1.09%	2.13%	21.29		Bush
ME	1968	54.75%	54.08%	-0.66%	1.13%	2.21%	27.80		Bush
MD	1000	57.00%	56.57%	-0.43%	1.58%	3.10%	39.18		Bush

DC	795	91.00%	90.91%	-0.09%	1.77%	3.48%	47.96		Bush
MT	640	39.76%	39.80%	0.04%	1.98%	3.87%	50.72		Kerry
OR	1064	51.20%	52.00%	0.80%	1.53%	3.00%	69.91		Kerry
HI	499	53.30%	54.55%	1.25%	2.24%	4.39%	71.10		Kerry
TX	1671	37.00%	38.38%	1.38%	1.22%	2.40%	87.10		Kerry

TN	1774	41.50%	43.00%	1.50%	1.19%	2.33%	89.68		Kerry
CA	1919	54.00%	55.56%	1.56%	1.14%	2.24%	91.35		Kerry
SD	1495	37.76%	39.39%	1.63%	1.29%	2.53%	89.65		Kerry
ND	649	34.00%	36.36%	2.36%	1.96%	3.85%	88.58		Kerry
KS	654	35.00%	37.37%	2.37%	1.96%	3.83%	88.76		Kerry

Avg	1450	49.18%	47.38%	-1.80%	1.42%	2.79%	21.67		Bush
Med	1507.5	49.23%	47.47%	-1.81%	1.29%	2.52%	6.66		Bush
									
						

[platinumman]

54. Yes, your data is incomplete

I have been questioning your work for weeks, now. Perhaps you should pay attention to replies instead of merely posting more and more of the same thing.

Yes, you are obviously not a polling expert. You don't appear to know much about them, and are jsut leading everybody up the garden path.

Those figures from SIMON/SCOOP only give figures state by state say nothing whatsover about what precincts were sampled, and you still have to justify your assumption that precincts were chosen randomly (contrary to normal commercial polling practice)

In answer to your questions:

You can use the numbers you used if you like, as long as you realise it is not telling us anything.
Mitofsky won't release the data because he is under contract to the people who paid him. It is also correct practice not to reveal which areas were sampled in order to avoid parties trying to distort the data. This is normally carried over for some months after the polling. It is just basic practice. There is nothing sinister about it.
I question the probability 'calculation' because, for the thousandth time, it is based on incomplete data: you don't know which precincts were measured.

Further:
1) Yes, you are using insufficient data to be able to make any sensible calculation

2) No, the Polling methodology was not flawed. It was probably basic methodology, which I have explained in past replies to you. As with most predictive polls, they were investigating not absolute numbers, but swing, so a 52% measurement in favour of Kerry could easily indicate a Bush win if that figure is less than the expected vote for Kerry in that sample

3) I would not say anything about the MOE as I don't know the sampling method. I'm sure it will probably turn out to be 2% or less. In other words, there is only a small chance that the measured value deviates from the actual IN THE PRECINCTS THAT WERE SAMPLED. But, as I said, we don't know which precincts were sampled.

[TruthIsAll]

62. INSUFFICIENT DATA? POLLING METHODOLOGY? MOE?

You say:
1) Yes, you are using insufficient data to be able to make any sensible calculation.

I say:
Insufficient data? 71,000 were sampled. I would assume the precincts sampled were chosen by experimental design to provide the most accurate results. The large number polled in what you call an "insufficient" data set is very telling in and of itself. The indiviual state MOE's were computed on that basis - and were close to 2.0% in the most important states. That is very telling.

The onus is NOT on my analysis. My analysis is correct, based on the data which has been downloaded and published. If you can get more data, I will be happy to use it. Until then, you are just spinning hypotheticals.

You say:
2) No, the Polling methodology was not flawed. It was probably basic methodology, which I have explained in past replies to you. As with most predictive polls, they were investigating not absolute numbers, but swing, so a 52% measurement in favor of Kerry could easily indicate a Bush win if that figure is less than the expected vote for Kerry in that sample.

I say:
If that's the case, why were most of the networks talking about Kerry leading the exit polls, yet at the same time lagging in the popular vote, when the bulk of the early votes were from the East (Kerry territory)? Why did Mitofsky let them use the polling data? After all, wasn't he still working for them? Did he let them because he knew the data was good and the pundits were justified in assuming that Kerry would win?

Let's assume that the 71,000 polled "early" represents 60%,70%,80% and 90% of the FINAL number to be polled, since I don't know what that number was. Let's also assume that Kerry was leading by 51-48% in the "early" polls.

This table shows what Bush would need of the remainder to be polled in order to attain 50%, 51% or 52% of the popular vote.

Note: I use 1000 as the final polling total in order to faciltate the calculation of the percentages of the remainder that Bush needed:

Early Kerry Bush Here is what Bush needs of the remainder:
Poll 51% 48% For52% Needs For51% Needs For50% Needs
60% 306 288 232 58.0% 222 55.5% 212 53.0%
70% 357 336 184 61.3% 174 58.0% 164 54.7%
80% 408 384 136 68.0% 126 63.0% 116 58.0%
90% 459 432 88 88.0% 78 78.0% 68 68.0%

Do you really believe that to reach 51% Bush would pull 55.5% of the remainder, assuming "only" 60% were polled?

Do you really believe that to reach 51% Bush would pull 58.0% of the remainder, assuming "only" 70% were polled?

Do you really believe that to reach 51% Bush would pull 63.0% of the remainder, assuming "only" 80% were polled?

Do you really believe that to reach 51% Bush would pull 78.0% of the remainder, assuming "only" 90% were polled?

You say:
3) I would not say anything about the MOE as I don't know the sampling method. I'm sure it will probably turn out to be 2% or less. In other words, there is only a small chance that the measured value deviates from the actual IN THE PRECINCTS THAT WERE SAMPLED. But, as I said, we don't know which precincts were sampled.

I say:
Well, I'm glad you will agree with me on that. But I have already calculated the MOE's for the early samples. The MOE's will only be reduced when, and if, the full data is released.

[jkd]

64. You are beating a dead horse

You are beating a dead horse. The numbers and the calculations are meaningless without Mitofsky's data. Sample surveys do have different results based on the quality of the data. You stated earlier that a poll is a poll. That seems to be your problem.

The pollster determines the margin of error based upon that data. Mr. Mitofsky prior to the election stated that the MOE for these state exit polls was + or - 4%. Many factors went into that determination. But we don't know what they were until Motofsky releases these considerations.

It's premature to say that the odds were 20 gazillion to 1. When we get the data and the recounts are completed then it should all fall into place.

[TruthIsAll]

65. If you believe 4%, you will believe anything.

You guys really don't know when to give up, do you?

Come on, you know as well as I, 4% is SHIT for an exit poll. Is that why they hired him? Exit poll samples are large, the individuals have already voted, the MOE is low.

That big an MOE? 4%
Why bother doing the poll?

At 8pm, Kerry was leading in the Exit polls nationwide by 51%-48%.
A total of 71,000 were already polled.
That's 63% of the final 113,000 count.

To reverse it to 51-48% Bush, he needed to win the remaining 42,000 by 56-44%.

That's a 15% turnaround from the first 71,000.
Even he's not that lucky.

The polls were RIGHT. The vote count was WRONG. Why won't you consider that? I never use Occam's razor. It's usually an excuse from doing your homework. But this time its right. The simplet explanation is that Bush stole it.

Get over it.
He has stolen it three times: '00,'02,'04.
He gets away with it each time.
Don't worry. He will again.
Maybe.

Why waste your time trying to convince me that I'm wrong?
I'm a nobody. Let me have my fun crunching numbers.
Why should that bother you?

Take your beef to Conyers.
Tell him to get the data from Mitofsky.
So we can put this all to rest.

[jkd]

66. Please answer

I've asked you perhaps 5 times to explain how the BYU and Mitofsky exits polls differed by 4%, if exit polls are so accurate. I'll keep asking.

[TruthIsAll]

67. The BYU exit poll in Utah? You are comparing it to Mitofsky?

Please, don't ever ask me that again.
BYU. Are you serious? One lousy poll?

One can always find an individual poll which will differ from another one.

WE ARE TALKING 50 STATES HERE, BUDDY - NOT ONE!
We have 50 states in the U.S., in case you haven't noticed.

Why don't you find something better to do with your time than to bother me?

I spend hours doing honest analysis and you keep popping up with pure BS.

The BYU poll. Sheesh.

I will ignore any post of yours from now on.

[jkd]

68. Yes, the Utah polls that Freeman cites

These are the Utah polls that Freeman says have a long history of accuracy in his paper. He used them to suggest the accuracy of the Mitofsky polls, but didn't seem to understand that they contradicted his conclusions.

[platinumman]

79. You have no idea

1) You 'assume the precincts sampled were chosen by experimental design to provide the most accurate results'. Well done. Good assumption. So you need to know the experimental design that Mitofsky followed to take into consideration the demands from the clients and the budget. Do you know this design? No, you don't. YOU HAVE
NO IDEA (sorry to shout, but you just don't seem to get the point.

2) You write: . "If you can get more data, I will be happy to use it. Until then, you are just spinning hypotheticals." EXACTLY!!!!! WELL DONE!!! You are at last getting the point!!!!
When you get more data from Mitofsky, you should use it. Until then, as you say, you are "spinning hypotheticals"

3) What do you mean "why did Mitofsky let them use the data?" It was their data. They could use it as they liked. It was for PREDICTION!!!!! Did any networks make any predictions based on early figures? I don't think so. I have no idea where you got the rest of these figures from, so I'll decline to argue with them.

4)Of course you haven't calculated the MOEs, because you have no idea what the design was, or what precincts were sampled. In any case, when we do know what precincts were sampled, the MOE will be irrelevant as we will simply be able to compare the measured poll data with the actual data. The real calculation is then whether that data could have provided us with a reasonable prediction for the state as a whole. That will be interesting when the data does come out.

[TruthIsAll]

47. Review this analysis from a senior research specialist (Ph.D.)

Ron Baiman, Ph.D. December 17, 2004.

Economist/Statistician Ron Baiman works as a senior research specialist at the Institute of Government and Public Affairs at the University of Illinois at Chicago, and teaches at the University of Chicago.



http://www.freepress.org/images/departments/997.pdf

[jkd]

50. Freeman and Baiman should stop citing the BYU polls.

Both Dr's. Freeman and Baiman cite the accuracy of the BYU student exit polls as suggesting that the professional Mitofsky polls must also be correct. Do they understand that Mitofsky had Kerry at 30.5% in Utah and BYU exit polls had him at 26.5%? The actual vote was 26%. The New Hampshire recount shows that Mitofsky missed even further there. So far it looks like the exit polls were in fact wrong and the actual vote was accurate.

[platinumman]

53. Who is Ron Baiman?

Ron Baiman is NOT an economist OR a statistician. He has written no statistical papers of any kind before this. He is NOT a member of any statistical organisation.

AND:
Take a look at the staff pages of the Institute of Government and Public Affairs at the University of Illinois at Chicago:

http://www.igpa.uiuc.edu/about/staff/default.htm

He is NOT on the staff there, although he has written a paper on subsistence wages for them.

Perhaps he works there part-time or something, but he is certainly NOT a senior research specialist.

Believe me, if the there were any sign that the exit polls proved or even indicated fraud, you'd have a million eager statisticians eager to be heard. When the full description of the reseach design is released, you can perhaps expect some papers to be forthcoming.

I do not discount the possibility of fraud, but to rely on the exit polls as proof is doing the case no good at all.

[platinumman]

55. Correction

In fairness, I should correct myself somewhat and say that Ron Baiman does appear to be an economist. He has contributed to some left-leaning books and publications, which I have no argument with. However, he certainly does not appear to be a statistician, and after considerable searching, I cannot find his name on any staff list of any university.
I have inquired among fellow statisticians, and his name rings no bells at all, I'm afraid.

[mod mom]

34. I always respect your posting TIA!

love the commentary. good points!

[Duncan]

35. Thanks again TIA.

I've not seen anyone even dent your logic.

Your work and work like yours will eventually be illuminated and will open millions of eyes. It burns me up that the FBI or someone has not subpoenaed the raw data. Have you put together a short, sweet, unassailable summary I can email to everyone?

[zimba]

40. Good work TIA

If this was the Ukraine or any other country, your data would have forced a revote by now.

[IndyPriest]

41. TIA, a little OT, but would you please weigh in over

here? The discussion at the end of this thread has gotten into the quesiton of whether past exit polls were accurate. I know you've commented on this, but don't remember what you said.
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=180307&mesg_id=180692&page=

[General Paranoia]

43. Gosh TIA seems to be stirring up concern

I haven't been here very long or made many posting but even I can discern how hard some parties are trying to dismiss TIA's findings. I wonder if someone could come up with a threat meter. I assume the higher the perceived threat the more hand waving dismissals we see. I wonder if it is an exponential per(hand wave dismissal) rather than just linear per (hand wave dismissal). Seems like TIA's posts are moving up from orange into the red.:wow:
:kick:

[TruthIsAll]

70. Give them credit. They are geniuses at the art of obfuscation.

They come swarming as soon as I post.

[Joe Chi Minh]

60. Would it perhaps be worthwhile, TIA,

enumerating at the end of your post, in a simple list, the falsehoods asserted by the mystery pollster, that you have identified, above?

The good or bad faith or at least the level of competency of the blogger in question might be more tellingly identified.

[Joe Chi Minh]

61. Sorry, TIA,

I've just read breadbox's posts and suspect the course he recommends you take, i.e. of a kind of amicable discussion and collaboration with him might prove more productive than the more adversarial one I suggested. Unless breadbox's assessment of him were to prove overly sanguine.

[TruthIsAll]

69. Nothing adversarial. I do my analysis. Let him do his.

If he wants to respond to my arguments, let him do so.
I post at DU.

I cannot believe that a polling expert could confuse the TV network's required 99.5% confidence level for calling the winner of an election with the 95% confidence level used by pollsters to determine a population mean. These are two totally different problems.

The problem here is NOT to declare a winner. The problem is simply to determine if the deviations from the exit polls for Bush were SIGNIFICANT, both in number and magnitude. We seek to derive a mathematical statement about the probability of this occurrence.

Using the 99.5% confidence criteria (or 3 standard deviations from the mean), effectively raises the MOE and lowers the number of states in which Bush tallies would exceed the MOE - when we all know that 16 did. The more state tallies which exceed the MOE, the stronger the circumstantial evidence of fraud. The MOE is based on THE STANDARD 95% CONFIDENCE LEVEL, regardless of the type of poll.

AGAIN, THESE ARE TWO COMPLETELY DIFFERENT PROBLEMS. OF COURSE, THOSE WHO ARE MATHEMATICALLY CHALLENGED WOULD NEVER KNOW OR APPRECIATE THE DIFFERENCE. BUT THOSE WHO KNOW THE MATH WILL NOT BE FOOLED. THAT'S WHY I RESPONDED ON DU.

[davidgmills]

74. Post Election Poll Ohio Now

TIA:

Your nemises on Mystery Pollster, of course, disagree with your analyses and even some who post on DU do as well.

The fact that there is no scientific, mathematical or statistical consensus on the validity of the exit polls has put the sitaution into a stalemate. It doesn't matter who is right or wrong on the mathematics and statistics at this point, because in order for those of us who might be inclined to take to the streets, a genuine clear and convincing consensus from the scientific, mathematical, and statistical community is needed.

I am a lawyer and have been for 27 years. I look at the situation just like I would if I were appealing a case. In law, there is something called the "harmless error" rule. It is what every lawyer fears who is appealing a case. It assumes there is no such thing as a perfect trial. It essentially says that in order to win on appeal, you must be able to convince the appellate court that there is a really good reason to believe that mistakes or errors made by the judge or jury likely caused the rendition of an improper verdict or decision.

Using this analogy to the present voting situation, if we are to convince the public at large that voting mistakes, error,and/or fraud caused Kerry to lose the election he actually won, we must have a clear and convincing consensus by the mathematical, statistical and scientific communities that such is the case. We simply do not have that now.

I believe such a consensus is possible, however. By re-polling Ohio now. I am told by your nemises on Mystery Pollster that it would take the polling of about 16,000 people in Ohio to basically be so accurate that most scientists, statisticians, and mathematicians could not reasonably disagree. They also believe that such a poll is not that unbearably expensive. I propose using three organizations to do one third of the polling each -- Zogby, Harris, and Gallup -- and have the results of the poll made public immediately.

I have asked MoveOn to help raise the money to fund such a poll. So far, I have not had any response. I believe such a poll would vindicate your work and shut up your critics for good. I believe such a poll is necessary now for the sake of the country and our democracy. Please help me turn this idea into reality.

[TruthIsAll]

63. Where did everybody go? I was just getting started. n/t

.

[raver]

72. We're here.

Continue. . . . .

Does n/t stand for "no/truth?"

Your math analysis makes sence so far. When can we expect the release of the actual raw data exit poll numbers to cross check your assumptions? Soon?

[berniew1]

73. Touchscreen machine fraud has been documented in swing state- explains swi

Systematic vote machine fraud, "Default to Bush" has been documented in swing states such as Ohio, Florida, and New Mexico- and some of this has been observed in other states as well.
http://www.flcv.com/fraudpat.html
Other types of systematic fraud have also been observed in many areas.
These appear to explain the swing to Republican seen by the difference in the exit polls and official vote totals.

This means that Bush or another default candidate would get not only the votes of those who didn't notice that the intended candidate didn't register, but any intentional non votes or accidental non votes or unsuccessfully completed votes. Quite an advantage. Under normal circumstances this would result in a swing as much as 2 to 4% in many areas. But since it is a matter of record that the default was hard to override and sometimes impossible as reported by the EIRS reports and acknowleged by poll workers, only the most watchful or purposeful voters may have been able to successfully vote for a chosen non-Republican candidate. Thus the swing was likely much larger for some machines and even more in some minority precincts that appear to have been targeted to produce high levels of misvotes based on the large number of EIRS reports in minority precincts. There were also a few cases where precincts had machines set to default for a minor party candidate, but none observed to default to Kerry. These patterns have been confirmed by analysts and computer experts in several counties of each of these states. Other similar default related patterns have also been observed in some of these areas, such as failure to register a presidential vote when a voter tried to vote straight party line. This would result in the presidential vote going to the other “unintended” party.

[17yroldtwins]

76. I don't know a lot about exit polls.

Did they poll for the gubernatorial and senate races at the same time?
If so, what was the accuracy level on those sort of races?

Can anyone point me in the direction for that info?

Thanks,
Attie

[lonestarnot]

77. Absolutely compelling circumstancial evidence!