47 State Exit Poll Analysis Confirms Swing Anomaly

Thursday, 11 November 2004, 1:45 pm
Article: Alastair Thompson

" In the 12 critical states (CO,FL,MI,MN,NE,NV,NH,NM,OH,PA,WI,IA) the average discrepancy was a 2.5% red shift (= total movement of 5.0%), nearly twice that in the safe states. "
- Jonathan Simon

- JUMP STRAIGHT TO THE NEW DATA

Introduction by Scoop Co-Editor Alastair Thompson

By the time of the close of polls at around 5pm EST on election day the buzz on the world wide web – including here at Scoop - was that Kerry had was a shoe in for election 2004. Slate Magazine and the Daily Kos had published the swing state exit polls before the polls had even closed. The news was very good for Kerry supporters.

According to the exit polls Kerry was showing a 1% popular vote margin over Bush. But more importantly he was shown leading by a nose in Florida and a solid 4% in Ohio. Because of the way the Electoral College system works this meant that he had almost certainly won.

The polls have significant sized samples in all states and ask actual voters who they actually voted for and so are traditionally very accurate.

As we now know they weren't very accurate once midnight came and went.

Or were they?

On November 4 (NZT – Nov. 3 EST) Scoop published Faun Otter: Vote Fraud - Exit Polls Vs Actuals. This was the first exit poll comparison analysis produced on the web it originated in the Democratic Underground, a forum website, for Democratic Party activism and the clubhouse for a lot of people doing grassroots research work.

Faun Otter's data - already immortalised in the Wikipedia (with a link to Scoop.co.nz) - showed swing states moving far further on average from their exit poll results than non-swing states after the polls closed. I.E. the actual result for these states was more at variance with the exit polls than it was in other states.

Alarm bells rang at this point because it has always been postulated that looking at exit poll results after a stolen election would be the best way to look for "general" evidence of voting fraud. By general evidence I mean evidence that suggests fraud has occurred – not proof that it has.

The reason this is so is that traditionally exit polls have been close to 2% accurate. Yet in the last three elections, 2000, 2002 and 2004 they haven't been. This years poll remarkably is almost precisely a re-run of 2000 with Ohio playing the part of Florida. In Florida in 2000 the exit polls showed Gore winning by 3%. In the middle of the night they were still counting and on this state alone hinged the entire election.

In 2004 just like in 2000 Fox news called Ohio to Bush before the counting had finished.

Because exit polls are such a good research tool for vote fraud analysis an organisation called PollWatch.org was even set up to conduct independent exit polls. By election time their efforts had been subsumed into the efforts of VerifiedVoting.org, a lobby organisation initiated by Stanford University Professor David Dill which signed up thousands of computer scientists and academics to a petition calling for auditable voting machines.

However the activists were caught off guard on election night.

The Official Exit Poll results – posted in real time on public websites - have some significant drawbacks. Unbeknownst to their readers CNBC, Fox News and CNN were constantly updating their exit poll databases to fit the final results. That is the statistics were fluid and were updated several times through the evening. By 2am in the morning on Nov 3, If you looked at the exit polls and the final results you would find the matched. For Ohio, for Florida, for everywhere. No story there people. Move on.

But as often seems to happen in these tortured times, something unexpected happened and so we can now tell you something close to the full story.'

The Washington post takes up the story:

Washington Post 11/4/2004:

"... a server at Edison/Mitofsky malfunctioned shortly before 11 p.m. The glitch prevented access to any exit poll results until technicians got a backup system operational at 1:33 a.m. yesterday.

The crash occurred barely minutes before the consortium was to update its exit polling with the results of later interviewing that found Bush with a one-point lead. Instead, journalists were left relying on preliminary exit poll results released at 8:15 p.m., which still showed Kerry ahead by three percentage points.

It was only after the polls had closed in most states and the vote count was well underway in the East that it became clear that Bush was in a stronger position in several key battlegrounds, including Ohio, than early exit polls suggested."

By 2am on Nov. 3 in the morning the publicly available exit poll results on the network news sites all changed. Activists still had the original results posted in blogs but they were no real comparison.

Which is why the following data study by Jonathan Simon of verifiedvoting.org is so remarkable.

As it turns out this study was only possible because of the computer crash reported by the Washington Post. While the boffins fiddled with their computers Simon – with a considerable degree of foresight - downloaded as much data as he could off the publicly available sites.

The revision number of this data is not known and the original data – from Edison - is now being sought by Scoop.co.nz in order to repeat this study with the full 4pm and 8pm data runs.

I conclude this introduction with some remarks from Chuck, who was commenting on Simon's results.

"Warren Mitofsky meanwhile says that he knew in the afternoon that his exit polls were off in nine states, but this does not sit well with me (I'd need to know how he would know at that point and, assuming he knew, why he would go ahead and promulgate them without caveat?).

Way too much work went into getting the exit polls right this time for me to just accept that they can't do as well as they were doing routinely in the 80s and 90s. It is not, like stained glass, a lost art."

Way too much indeed.

- Alastair Thompson Scoop Co Editor Thursday, 11 November 2004

I examined the discrepancies between the actual vote tabulations as reported and the Edison/Mitofsky exit poll results in 47 states, incl. D.C. (in 4 states—New Jersey, New York, North Carolina,Virginia—I did not have early exit poll results available, and the later results had already been amended to reflect input of actual vote totals, which rendered them corrupt as exit polls and useless for the purpose of checking the veracity of actual vote totals).

I noticed an overall red shift (to Bush) across the spectrum of states, but the shift was significantly nonuniform.

Having divided the 47 states examined into two groups, 35 noncritical states and 12 critical or suspect states (Nebraska included because of ES&S control and prior anomalies even though not a battleground state).

I calculated that the average discrepancy in the 35 safe states was a +1.4% red shift, that is the average of the vote totals in each state was 1.4% more favorable to Bush than what the exit polls predicted (= total movement of 2.8%).

In the 12 critical states (CO,FL,MI,MN,NE,NV,NH,NM,OH,PA,WI,IA) the average discrepancy was a 2.5% red shift (= total movement of 5.0%), nearly twice that in the safe states. This in spite of the fact that the average sample size in the critical states was nearly twice that in the noncritical states and should have produced significantly more accurate results.

Further, assuming a 3% margin of error and 95% confidence interval for each state poll (the standard Mitofksy protocol, but a conservative assumption here, since the sample sizes were significantly increased in critical states), the red shift exceeded the margin of error in 4 of the 12 critical states (and equalled it in a fifth).

The chance of this occurring in 4 of the 12 states in the absence of "mistabulation" can be computed using a simple probability equation and is approximately 0.002 or one in five-hundred. It's a relatively crude analysis and better analysis would have to wait on more complete data, but basically what it's telling us is that we can say with 99.8% certainty that "mistabulation" played some significant role in this election.

Critical States (12)

DATA DESCRIPTION
Exit Poll Data Bush% then Kerry%, # of respondents, then time of poll ET, and "Red" Shift%
Note: Red Shift = <(Btab% - Bep%) + (Kep% - Ktab)>/2 tab= tabulated vote, ep=exit poll

The number is positive with net movement toward Bush, negative (blue shift) with net movement toward Kerry. I'll take Florida (early) as an example:

Exit Poll % : B=49.8% K=49.7%
Tab (99% precincts) B=52% K=47%
Red Shift: <(52% - 49.8%) + (49.7% - 47%)>/2 = (2.2% + 2.7%)/2 = +2.5%

I'm aware that I've played fast and loose with significant figures; a more refined analysis would get at least one more sig fig out of the tabulated.

Important Note: Because of rolling updates, some states may have been relatively pure by the time this data was collected. at 12:20-12:25, while others may already have been slightly corrupted. My guess is that most of these states were still OK at these times. *NOTE SCOOP IS SEEKING A COPY OF THE ORIGINAL DATA FROM 4PM AND 8PM ON ELECTION NIGHT – WATCH THIS SPACE*

Non-Critical States (35)

Again data order is Exit Poll Data, Bush% then Kerry%, # of respondents, then time of poll ET, and "Red" Shift%

FINDING: 35 (NonCritical) State Average Red Shift +1.4

The following state data was obtained too late and highly contaminated with actual tabulation results:

Important Points To Note: The average sample size for the critical states is 2109, for the noncritical states 1192, roughly half the size. All else being equal, higher sample size correlates with smaller margin of error, greater accuracy. And yet: the average Red Shift in the noncritical states is 1.4%, but in the critical states it is 2.5%.

Why? Hard evidence must be found to account for this egregious statistical pattern and to prove the actual fraud. This preliminary analysis is fairly crude, and there are probably other, more telling ways to slice and dice the data. Also, note that there were three exit poll "sweeps" which led to updates around 4 pm, 8 pm, 12 midnight, following which the "exit polls" were really contaminated by tabulated data and of no use at all as a check mechanism.

I caught mostly the third sweep here (c. 12:20 am); my guess is the second sweep would show even more dramatic discrepancies. While the first sweep has come under skeptical attack for having a too great female (pro-Kerry) weighting (the figure I heard was 58%/42% female), it is at least plausible to me that morning/daytime voters would be disproportionately female because of work schedules, so it may well be just one of the specious rationalizations behind a very determined coverup (we just don't know).

—Jonathan Simon
Thursday, 11 November 2004

ENDS

© Scoop Media

Alastair Thompson

Scoop Publisher

Alastair Thompson is the co-founder of Scoop. He is of Scottish and Irish extraction and from Wellington, New Zealand. Alastair has 24 years experience in the media, at the Dominion, National Business Review, North & South magazine, Straight Furrow newspaper and online since 1997. He is the winner of several journalism awards for business and investigative work.

Contact Alastair Thompson

• Linkedin
• Twitter - @althecat
• Facebook
• Email - alastair@scoop.co.nz