Anyone know why exit poll numbers were adjusted late in the evening?

[democraticinsurgent]

I have yet to see an explanation of why the exit poll projections which were so favorable to Kerry at 6pm were changed later in the night. Were these updated by the Mitofsky-Edison organization due to new data that came in at the end of the day? Or did the networks adjust them to match the official results?

Much is being made of the exit polls being "wrong" in the major media. Many of us here instead feel the exit polls were right and there is monkey business with the reported vote totals.

But my understanding is that the "adjusted" exit poll results were close to the "actuals".

So I'm still trying to understand. Did the exit polls get it wrong at 6pm and then measure a late Bush surge which led to revised/final exit poll projections that more closely match the official results?

If that's the case, then the exit polls did their job. However, this does not seem to be the case given the hue and cry over exit polling in the mainstream media.

Anyone know the real story here?

[BJ]

1. Here's my best guess.

Christian Reconstructionists lie

I've read it somewhere that "Christian Reconstructionists" feel that it is permissible to lie to an unbeliever if they feel it will further the work of the church.

But that's just my supposition

[On the Road]

2. I Believe It Was Because Precincts Were Originally Weighted by

the expected turnout.

You can't simply add all the raw numbers together from different places and get an accurate sample. A exit poll worker at a tiny polling place, for example, is likely to sample a much higher proportion of the voters than a team of people at a large urban polling place.

So the results from each location have to be weighted by how large a percentage of the total they're expected to be. It's probably based on 2000 results modified by changes in population and registered voters. And until the results are in, you never know what the real weightings should be.

Usually the process works pretty well. But this was not a normal election. For whatever reasons, the increase in turnout among Bush supporters broke every record and trend in predicting elections (see below). That meant that all the old weightings had to be replaced.

I don't rule out election fraud, but the change in exit polls does not prove it. The Republican turnout was so deep and wide it cannot all attributed to that. And under those conditions, reweighting the exit polls could not be avoided.

--------------------------------

Florida
Increases in Turnout for Bush

County	      % Incr
Osceola	        +64%
Dixie	        +64%
Sumter	        +63%
Flagler	        +56%
Columbia        +53%
Levy	        +52%
Pasco	        +50%
Wakulla	        +50%
St. Johns	+50%
Gilchrist	+50%
Lake	        +49%
Clay	        +48%
Washington	+48%
Marion	        +47%
Lafayette	+47%
Liberty        	+46%
Union        	+46%
Nassau	        +45%
Duval        	+44%
Walton        	+44%
Seminole	+44%
Santa Rosa	+43%
Orange        	+42%
Franklin	+42%
Manatee        	+40%
Alachua        	+40%
Bradford	+40%
Suwannee	+39%
Palm Beach	+39%
Bay        	+38%
Collier        	+38%
Okeechobee	+38%
Madison	        +38%
Baker	        +38%
Broward        	+37%
Polk        	+37%
St. Lucie	+37%
Putnam	        +36%
Lee	        +36%
Gulf	        +36%
Volusia        	+36%
Hillsborough	+35%
Taylor	        +35%
Hardee        	+34%
Okaloosa	+34%
Jefferson	+33%
Citrus	        +33%
Brevard        	+33%
Jackson	        +33%
Leon	        +32%
Gadsden	        +32%
Calhoun	        +32%
Hernando	+31%
Hamilton	+30%
DeSoto	        +30%
Indian River	+29%
Holmes	        +29%
Highlands	+28%
Escambia	+28%
Sarasota	+26%
Charlotte	+25%
Miami-Dade	+24%
Martin        	+22%
Pinellas	+22%
Hendry	        +21%
Monroe	        +21%
Glades        	 +8%

[tngledwebb]

7. Hmmm.

Even this ska-tistical anomaly should have been accounted for empirically through previous election studies and polls.
Or is this some new science only relevant to the quite unexpectedly huge Bush voter turnout? Plus the Bush supporters I know have quite a lot of trouble keeping quiet about their crazy ideas, and also, at least in my neighborhood, I witnessed quite the opposite in turnout numbers.

[On the Road]

9. Turnout Varied Widely by Neighborhood

but in general it was much heavier than past elections.

[blurp]

17. Can't be accounted for. You can only estimate turnout.

Even this ska-tistical anomaly should have been accounted for empirically through previous election studies and polls.

That's exactly why the numbers were wrong.

What they do is interview a certain number of voters from various precincts, look at the percentages for Bush/Kerry, etc, and multiply by the predicted turnout.

If you underestimate the turnout, you underestimate the number of votes.

[blurp]

8. Yep. They weighted the results based on PAST elections. Oops.

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

"How do you select sample precincts?"

"The polling places were selected as a stratified probability sample of each state. The purpose of stratification is to group together precincts with similar vote characteristics. A recent past election was used to identify all the precincts as they existed for that election. The total vote in each precinct and the partisan division of the vote from this past race are used for the stratification. In addition, counties are used for stratifying the precincts. The total vote also is used to determine the probability of selection. Each voter in a state has approximately the same chance of being selected in the sample."

It appears the sample was not completely random.

[On the Road]

10. It's Not Really an "Oops"

because there's no way to weight correctly without knowing how many people voted in which precincts. That information doesn't exist until the results are reported. So unless the exit pollsters can measure turnout, any early exit polls MUST be based on historical estimates.

Most of the time the estimates work fine. This time they didn't. Come to think of it, maybe that was the problem in 2002 as well.

[blurp]

13. So the polls failed because turnout %s changed. Okay.

I think I understand what you mean.

They take samples of people from various precincts and multiply by the estimated turnout for that precinct to get the numbers.

So if they underestimate turnout, they also underestimate votes.

This must mean Repub won the turnout battle.

That's another reason Dean should have been the nominee, I guess.

[On the Road]

20. Yes, the Republicans Appear to Have Won the Turnout Battle

and under these conditions, there was no way the exit polls could have been right.

I'm not at all willing to concede that no fraud took place, just that the exit polls aren't the evidence a lot of people think they are. There are many ways to commit fraud. Republican votes, for example, could have been manufactured out of thin air. This could be done on any voting system: extra optical scan ballots or punchcards could have been placed in stacks of genuine ballots.

For that reason, it would be helpful if there were a way to ascertain whether the number of votes recorded are equal to the number of people who actually voted.

[NewYorkerfromMass]

3. I believe they looked at respondents and determined they were imbalanced

The demographics werenot representative. The NYTimes has an article on it today.

[Young_Monk]

4. Why?

Everything down the memory hole...

History is written by the victors, and the internet has just sped up that whole process.

[NewYorkerfromMass]

6. Idunno- "Report Says Problems Led to Skewed Surveying Data"

The new $10 million polling system used by many news organizations to predict the outcome of the presidential race had a number of problems that led to the early erroneous impression that John Kerry was heading for victory, according to a report prepared by the system's architects.

The report, written by Joe Lenski and Warren Mitofsky and obtained by The New York Times, details systemic glitches that skewed the data in ways of which several news organizations, who paid tens of thousands of dollars for the service, were not aware.

In some cases, the report said, survey takers could not get close enough to the polls to collect adequate samples of voters opinion. They were often stopped by legal barriers devised to keep people electioneering - not necessarily bona fide poll canvassers - away from voters.

The report also theorized that the poll results more frequently overstated support for Mr. Kerry than for President Bush because the Democratic nominee's supporters were more open to pollsters. Whatever the case, according to the report, the surveys had the biggest partisan skew since at least 1988, the earliest election the report tracked.....

http://www.nytimes.com/2004/11/05/politics/campaign/05poll.html?pagewanted=print&position=

[On the Road]

11. That Last Item Might be True

when I canvassed and got out the vote for Kerry, the Bush supporters I met were very defensive and didn't want to talk. All it takes is one out of 20 voters to skew the polls by 5%.

[NewYorkerfromMass]

19. Of course they were. I knew there was a silent Bush support

which was too embarrassed to admit it publicly.

[tngledwebb]

5. If the NYT says so,

It must be true. Thanks for clearing everything up.

[TruthIsAll]

12. The early EXIT polls were correct. Here is preliminary proof of fraud.

Historically, Exit polls are accurate within 2%

The deviation is the difference in percent between the first exit polls and the vote. +
2B means the change was + 2 for Bush

All these states were accurate - they all have paper trails
...Deviation
AZ 0
LA +2B
MI 0
IA +1B
ME 0 <no BBV
NV 0 < paper trail
MO -1B
IL 0 paper ballot

Now look at the states that in play where there was NO paper trail.
They ALL had Bush do much better than the Exit polls indicated.
They wre all WAY outside the MoE.

AND THEY ALL WERE SKEWED TO BUSH!

Any deviations above 5% have a probability of 0 to 1%.
Deviation = 4% has a probability of 2%.
Deviation = 3% has a probability of 4%

This is a clear circumstantial case for fraud.

WI +4B
PA +5B
OH +6B
FL +7B
MN +7B
NH +15B
NC +9B
CO +4B
NM +3B

[blurp]

14. Uh, the people doing the polling say +/-4%. How can you say 2%?

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

"What is the Margin of Error for an exit poll?"

"Every number estimated from a sample may depart from the official vote count. The difference between a sample result and the number one would get if everyone who cast a vote was interviewed in exactly the same way is called the sampling error. That does not mean the sample result is wrong. Instead, it refers to the potential error due to sampling. The margin of error for a 95% confidence interval is about +/- 3% for a typical characteristic from the national exit poll and +/-4% for a typical state exit poll. Characteristics that are more concentrated in a few polling places, such as race, have larger sampling errors. Other nonsampling factors may increase the total error. "

[rinsd]

16. And that is after figuring in actual turnout (nt)

[blurp]

18. Right. So if you misestimate the turnout, it's even bigger.

[RevCheesehead]

15. Gee, could it be the media wanted to defend their skewed polling?

How simple it is to say, oops, we were wrong with our exit polls... and this just goes to show you that OUR PREVIOUS POLLS were more accurate than Zogby, the most reliable pollster...

I just don't buy it that Repug turnout was higher than Dems. No way - especially with an approval rating below 50%. (and remember, some of those same polls started moving *'s approval ratings higher as the election drew near.)

[MISSDem]

21. Please explain this to me?

How could the exit polls be so wrong?

[blurp]

24. It was the extrapolation that was wrong

What they apparently did was take the percentages of Kerry voters / Bush voters, etc and multiply by turnout estimated from previous elections.

The trouble with this is that you undercount areas that have a higher turnout this year than last year.

See, they don't know how many people will show up before hand to vote, so they use previous election turnout numbers to guess. If, for example, 1000 people showed up last year at a certain precinct and 50% vote Kerry THIS year, you assume 500 people will vote Kerry in that precinct.

The trouble occurs when turnout increases or decreases unevenly at the various precincts.

Suppose fewer people than the 1000 estimated show up to vote for Kerry. In that case, estimating 500 votes for Kerry is an overestimate.

Likewise, if 2000 show up, then 500 votes for Kerry is an underestimate.

The trouble in this election seems to have been with underestimating turnout in more conservative areas and so underestimating votes for Bush.

[slackmaster]

22. My cynical opinion about exit polls and reporting

Follow the money.

Any manipulation of exit poll results was done for the purpose of keeping as many people as possible glued to the boob tube.

The only poll that really matters is the vote count. That and nothing else should be our focus. Exit polls are irrelevant, inadmissible fluff.

[bloom]

23. I think it is too convenient that the polls went from Kerry up 5 points

to Bush up by 2.5 points to fit the number of votes that were supposedly registered.

If this were a "third world" country we wouldn't buy it.