OK, MYSTERY POLLSTER: REFUTE THE MATH, IF YOU CAN.

[TruthIsAll]

You have been trying your best to pooh-pooh Freeman, Berkeley, Paulos and others.

Now it's time to put up or shut up.
Show us your probability model, if you have one.
Time to fish or cut bait.

THESE PROBABILITIES HAVE NOT BEEN REFUTED; IN FACT, THEY HAVE BEEN CONFIRMED BY OTHER MATHEMATICIANS HERE AT DU.

These are the probabilities that Bush's tallies could have deviated as they did from the exit polls based on the MOE and calculated using the binomial distribution:

1) If you assume the calculated MOE, based on sample size (as for standard, non-exit polls), then the ODDS of Bush's vote tallies exceeding the exit poll MOE in at least 16 states is
1 out of 13.5 TRILLION.

2) If you assume a 2% MOE (more likely for exit polls), then the PROBABILITY of Bush's vote tallies exceeding the 2% Exit Poll MOE in at least 23 states is ZERO. YOU CAN'T COMPUTE THE ODDS, UNLESS YOU HAVE FIGURED OUT A WAY TO DIVIDE BY ZERO.

If N = number of states exceeding the MOE, then the probability is:
Prob = 1-BINOMDIST(N-1,51, MOE/2, TRUE)

For (1) and (2) above:
P(1)=1-BINOMDIST(15,51,.015,TRUE);the odds are 1/P(1)=1/13.5 trillion

P(2)= 1-BINOMDIST(22,51, .01, TRUE) ; the odds are 1/P(2) = 1/0 =???

It's as simple as that. Try it yourself in Excel.

[Igel]

1. I have a simple question.

Everybody's 100% sure that the exit polls are accurate.

So, what's the margin of error at the 100% (oh, ok, try 99.9%) confidence level?

[papau]

2. The number in each state exit poll is given - so it's an easy calc - but

just compare National polls with 500 folks, to exit polls with 3000 in a state.

[TruthIsAll]

34. Papau, national polls usually sample 1000-1250; state polls 400-600

The MOE for national polls is around 3.0%
The MOE for state polls is around 4.0%

[papau]

38. The State numbers I saw posted (I DU mailed to you) had 2000 or more

in the exit polls for any given state.

I agree the before election polls were as you state.

But I believe the total count in ALL the exit polls nationwide was between 55000 and 75000.

That is why exit polls are thought to be so accurate!

In any case - the DU mail I sent is below

As posted at http://exitpollz.org
(THANKS to [e-mail] )

The "Original" 2004 election FULL end-of-day polls that were available on media sites on Nov.2 from about 7 p.m as the "exit-polls", and before the later 1.30 a.m. updates made to fit the "real vote".

Colorado Presidential exit-poll (html)

Number of respondents 2505
Time of Weighting: 11/02/2004 9:26 pm
Kerry
Bush
Expected result:
49.1%
50.9%
Declared result:
47.3%
52.7%


Florida Presidential exit-poll
Number of respondents 2826
Time of Weighting: not indicated
Figures published on CNN indicating that they were "weighted".
Kerry
Bush
Expected result:
49.9%
50.1%
Declared result:
48.7%
51.3%


· Ohio Presidential exit-poll
Number of respondents 1963
Time of Publication: 11/02/2004 7:32pm
Time of Weighting: not indicated.
Figures published on CNN indicating that they were weighted.
Kerry
Bush
Expected result:
52.1%
47.9%
Declared result:
48.8%
51.2%


Compare how accurate the senatorial exit polls below were …

· Colorado Senatorial exit-poll
Number of respondents 2491
Time of Weighting: 11/02/2004 9:26 pm
Salz (D)
Coor (R.)
Expected result:
52.2%
47.8%
Declared result:
52.2%
47.8%
Discrepancy: None! Spot on!

Wisconsin Senatorial exit-poll
Number of respondents 2157
Time of Weighting: 11/02/2004 8:55 pm
Fein (D)
Mich (R.)
Expected result:
56.0%
44.0%
Declared result:
56.6%
44.4%
Discrepancy: Almost none! + 0.6% to Fein(D)

[MrUnderhill]

3. They've been "refuted" a number of times..

...just not "disproved".


As I said the first time you asked.... there isn't necessarily anything wrong with your MATH (though you've spotted a few errors and corrected for them) the original point wasn't 10 Trillion:1 vs. 13 Trillion:1 - it was that it's a really big number and I agree that the math works out that way.

What you haven't demonstrated is any reason to believe that your mathematical model accurately represents reality. IF the only possible error were statistical sampling error based on sample size... you're math would be adequate. But that isn't even close to true.

[TruthIsAll]

4. You agree the math is correct. So you are left with grasping at straws.

So enlighten us. What are the possible causes of error?

Give us your explanations.
Will you site exit polls in Ukraine?
Or Germany?
Or Venezuela?
They were all correct to within 1%.

Was it the lack of experience of the exit pollsters - Mitofsky et al?
Hardly.

Was it your faith in Bush, that he would never stoop so low as to try to steal the election a SECOND time?

What is it?
You offer no legitimate rationale for your case.

[donkeyotay]

8. He's referring to the cover legend that was in place on Nov. 3rd

And the fact that the legend was being beaten into dense skulls so soon after the election should be a clue that it's need had been anticipated. The legend goes that the brave, independent, wily Republican has no time for pollsters. That this noble warrior is a separate class of person, one who does not speak to pollsters. (Of course, if you've ever been around any of these obnoxious crows, you'd immediately suspect this story.)

Just as when Wellstone's plane crash, when the legend that it was caused by ice was being beamed into America's brain before his plane had hit the ground (even though the subsequent NTSB reports it does not know the cause of the crash)the mere existence of such a ready excuse should cause one to be suspicious... unless you are one of the people being flattered by the legend.

Carry on. The Truth will set us free.

[MrUnderhill]

12. Nope...

I don't buy the notion that "republicans don't talk to pollsters", though I'd place the likelihood of that at greater than 1% as an explanation (IOW... FAR more likely than 13T:1).

We saw the same types of explanations for pre-election polling. "younger voters use only cell phones and are therefore less likely to be reached by a pollster - and they're going to vote like they never voted before" was the mantra. Along with "late deciders ALWAYS break toward the challenger". These dramatic oversimplifications did a disservice to people trying to interpret polls (Truth here just operated under the assumption that any "undecided" in a poll was automatically a vote for Kerry).

[Name removed]

28. Deleted message

Message removed by moderator. Click here to review the message board rules.

[MrUnderhill]

33. I didn't "lie"

At worst, I was wrong. In reality I was exaggerating. It was my interpretation of how you could have days where not a SINGLE poll showed Kerry ahead, yet you would magically convert it to a virtual Kerry lock on the election. They don't break 80-20 either. And it isn't correct to state that is "widely accepted by professional pollsters" but I refuse to accuse you of LYING in all caps. You're simply wrong. You again accept some limited previous history as a magical force that cannot be stopped - without ever understanding WHY people break one way or the other and whether that might or might no hold true this time.


And you weren't "right". Nobody in the party was talking about "99.7% chance of victory", we knew this was going to be close.

You had days where every single poll showed Kerry behind by averages as high as 6% and you were STILL running numbers with incredible Kerry victory percentages.

And I doubt you've done a "complete analysis" in your life. You take ONE tiny little piece of a puzzle and blow it up to be THE WHOLE puzzle and "analyze" it six ways from Sunday and marvel at how anyone could disagree with you.

[Name removed]

37. Deleted message

Message removed by moderator. Click here to review the message board rules.

[MrUnderhill]

40. Yes. I'm a Democrat.

I'm just not currently under the influence of any mind-altering medications.

I'm not saying I'm certain Bush won.... I'm not. I'm just saying that mathematical games showing it was IMPOSSIBLE that he won are ridiculous.

[Name removed]

43. Deleted message

Message removed by moderator. Click here to review the message board rules.

[MrUnderhill]

54. I don't need to "defend" anyone.

You're making a false analogy.

Like I said before. Imagine a die-hard Democrat posts that Bush had the CIA kill Chirac and replace him with a mole that would support the next war. When I say it's ridiculous... I'm not "defending Bush" just because I disagree with silly thoughts.

As I said a number of times. Kerry may very well have won - I hope he did. Nothing would help the party more than to lose and then have the Ohio recount show 500,000 miscounted ballots. It would be FAR better than simply winning a fair election.

I'm NOT saying there wasn't cheating. I'm saying I don't KNOW, and disagree with silly CT calculations that pretend there is a mathematical certainty that it happened.

This is a mis-application of mathematics. The problem has not been set up properly. The fact that the calculator can give an answer does not mean the answer bears ANY relation to reality. This one doesn't.

[emcguffie]

77. are you a --

mathematician or statistician?

Because that business about Repubs not talking to pollsters doesn't wash. Some would, some wouldn't. In another state, or another precinct, maybe more would than wouldn't. Maybe the Dems wouldn't.

Isn't that the reason this polling business works at all, because the anomalies cancel themselves out, because some would go one way and some would go the other? Just like a coin will fall one way sometimes and the other way sometimes? How often does a quarter land on the same side a thousand times in a row? And this is better than a pre-election poll, because they're right there, and they've just voted?

All Republicans do not have the same propensity to speak to an exit poller. That cannot be. Nor do all Dems. Or all independents. In fact, the Dems are probably more likely to be in a hurry to get back to work, so they won't get fired. Or get fined $1 per minute for being late to pick up child at daycare center.

[MrUnderhill]

92. Hmmmm. Yes and no.

I majored in Physics, not mathematics... but why split hairs? You have to take enough math that you're only a couple classes away from a degree in Math as well.

On the other hand... there isn't much "math" being done here... it's mostly setting up a simple formula in Excel (and making a couple mistakes there if we're taking a number too small for Excel to be "zero", but that's splitting hairs).

On the third hand... I don't do that for a living and I'm a lot of years removed from those courses and don't remember THAT much physics. Most of the mathematics I use now are limited to banking purposes. You want to talk about various methods of amortizing loan payment... I'm your man! :-)

But as I said... I don't buy the "Republicans don't talk to pollsters" either. I'd need to see some compelling data before I would buy that.

But in reply to your second point. No. The anomalies do not "cancel themselves out". That's why one polling firm will differ so much from the next (in some cases even when using the SAME raw data). NO polling firm reports the actual raw figures... there's always some "massaging" done. The intent is (usually) to end up with MORE accurate results... but if the assumptions are wrong, the final reported poll will be as well.

You are correct that an exit poll at least gives you actual voters (assuming someone isn't playing games with your audience), but it's important to remember that this also causes potential problems. Remember... different precincts in the same county can have DRAMATICALLY different results... it's the PICKING of WHICH precincts to run the polling in that is so important. They only choose a handful of precincts in each state and there's a LOT of science that goes in to trying to pick representative precincts... but this isn't without possible error, because they need to make SOME assumptions about where the voters will turn out. Where has the voting population shifted in the last four years? Where was the heavy registration activity? Will those new registrees actually turn out?

If they made a single mistake in their presumptions, it could throw off ALL of their numbers - and in the same direction. These mistakes aren't that hard to make. Our party strategists have said that OUR turnout was what we had targeted. In Ohio and Florida, the people we needed to vote actually showed up and we hit our targets... but (they say) the Republicans did a better job - we misjudged their turnout in certain demographic areas. If, as they claim, this came from certain types of voters/areas, it isn't hard to imagine that the exit pollsters operated under faulty assumptions as well. These "exurbs" that the keep talking about are relatively new demographic areas caused by skyrocketing real estate values... maybe we misunderstood how many of them would vote.

[KerryOn]

93. Single Mistake Could Throw Of All Numbers?

I live in a neigborhood that is primarily made up of Republicans, so if they had a polling place set up in my precinct it could in fact swing the exit poll towards Bush.

However, no one will ever convince me that a single mistake could throw off ALL of their numbers in the same direction. The statistical odds are against this happening in vertualy every state.

Impossible.

[MrUnderhill]

94. Yep. But it depends on the mistake.

A single mistake of choosing YOUR precinct vs. the one next to it would throw off very little... unless perhaps your precinct had changed unexpectedly (say... it was heavily Republican but the pollsters didn't realize that a high percentage of you work for the same company and had recently been laid off). Even then it would be no big deal.

But many of the decisions they have to make affect the polling nationally. Say the far suburbs vote more heavily than expected and the pollsters believe us when we say that we've dramatically increased inner-city voter registrations and are expecting record turnout. If they misjudge the proportion of the vote coming from these areas, they might have picked too many urban precincts and too few far-suburb precincts... over-representing (slightly) Democrats and under-representing Republicans. That could be nation-wide.

I'm not saying that is what happened... just giving an example.

[KerryOn]

95. You made a good point

I know that here in Ohio Bush carried most of the country folks, and Kerry had more of the city areas, with the exception of Fanklin county.

I would be curious to know weather this was true nation wide.

[alittlelark]

63. Could you please post your analysis

TIA posts his #'s - perhaps you should do the same if you wish to 'sway' US.

[MrUnderhill]

67. That's a false dichotomy.

I'm saying that the "analysis" is nothing fo the sort. It isn't a valid way of proving that the election was stolen.

I don't have to have an alternative way of proving it to demonstrate that he isn't taking everything in to account. I certainly don't need a mathematical proof that Bush won if I don't WANT him to have won.

I'm saying there IS NO WAY to mathematically demonstrate this the way he wants to. Bush either DID or DID NOT cheat. We need to prove it from actual evidence. Not some speculation that exit polls are always perfect and impossible to rig so there's no chance they were wrong.

[alittlelark]

69. uh, yeah........

<putting on ignore>

[MrUnderhill]

70. Oh NO!

Not "ignore"!

But what about my fragile little ego?

I MUST have YOUR input in order to SURVIVE!

Please PLEASE don't leave me all alone in this big bad world!

Boy... I guess you showed me.

[Wordie]

44. Please read and respond to my post below. n/t

[RaulVB]

36. Yes, they are "pure", "relentless," "courageous..."

"The last line of defense..."

"They will protect us from ourselves..."

[MrUnderhill]

11. Grasping at straws?

Please. I don't have ANY faith in Bush, but I don't have to accept every conspiracy theory that walks down the street because of that. If you posit that Bush had the CIA bomb trains in Spain so that he could raise terror awareness for our election, my skepticism has nothing to do with any "faith" in Bush.

So here's a few:

1) You treat "exit poll" as if it's some magical word that ensures reliability. It is not. Exit polls conducted with different methodologies by different polling groups for different reasons (issues vs. results) don't become magically perfect because both use the word "exit". It's like comparing Zogby to some far-right firm and saying they're both accurate because they both do "telephone polling".

2) Exit polls have not been uniformly reliable in predicting US races at many levels. "Calling" the race isn't there primary purpose (as it is when we run them to confirm elections overseas), and the more you concentrate on one priority, the less of a handle you have on the others.

3) You haven't addressed (to me - I do not intend to read hundreds of posts on your daily versions of this thread to see if you resolved it with others) that some of the numbers in those exit polls don't make any sense. When I asked you how believable it was that 57% of the electorate in some of these swing states were women you just said "exit polls ARE accurate".

4) Both parties have speculated for YEARS about who votes when (it's even in some West Wing episodes). "Democrats vote early", or "Republicans vote after work", or "people with jobs can't wait in lines" or dozens of other possibilities. It's certainly more than "13Trillion to one" likely that the people who voted in the morning and early afternoon were not indicative of the people who voted in the evening. The reorts could even have worked against us (though I wouldn't expect it to work that way) by envigorating the opposition's attempts to turn people out and making our team think it was already in the bag (I admit that I loosened up substantially in the early afternoon and stopped worrying about the election... until around 9pm when it all fell apart).

5) Possible fraud the OTHER way. It's a heck of a lot easier to cook the exit polls than it is to cook the returns. All I need is foreknowledge of WHERE the polls will be run and a few dozen people willing to twist them my way. Get some early reports that I'm winning and pour cold water on the other guy's turnout. Heck... you don't even need to do that... just convince a couple guys INSIDE the polling firm to cook the books slightly (they never use "raw" data, just like other polling firms there is some "massaging" that goes on in an attempt to make them MORE accurate). A minor change here or there would be HUNDREDS of times easier in the relatively unsecured environment than it would be for right-wingers to cook the books in secure Democratic counties.

6) Recounts. So far, where they've recounted, the actual returns match the reported returns FAR more than they do the exit polls. When they hand-count Ohio, they're either going to find small discrepancies (disproving your thesis), or they;re going to find that there are 250,000+ votes that the machines lied about. We'll know pretty soon.... but I suspect you won't change your mind - because those magical exit polls are divine.

7) Florida in 2000 was COMPLETELY counted by hand... and while we found plenty of evidence that Gore actually WON... NOTHING indicated the exit polls were even CLOSE. There wasn't anything left to hide there - all that is left are disagreements on HOW/WHETHER to count a very small number of ballots. There's simply no way the exit polls were right there.

8) Poor polling is FAR more likely than 13Trillion to one. They only pick a TINY TINY handful of precincts. Picking them poorly could easily skew the exit results. And if one or two firms picked the "representative" precincts the error could have been uniformly in one direction. Say, for instance, they misjudged turnout in growing areas (or misjudged the rate of growth) and picked too many urban precincts and not enough "exurban" ones. A very small error could throw off the results by 1-2% across the board (and all in one direction).

That's just off the cuff. I'd bet I could come up with a dozen more.

Note... I'm NOT saying that Kerry didn't win. I still hold out hope that this is a possibility. I simply laugh at suggestions that the statistical likelihood that he won approach a certainty. I'd think 10% at this point would be incredibly good chances.

You were similarly off on your pre-election statistics. Again - in the assumptions, not necessarily in the working of the math. You make the error my old math students always made. When doign a "word problem", the key is setting the formula up correctly... not in actually working the formula (that's the easy part).

[cdp]

13. I agree.

"I simply laugh at suggestions that the statistical likelihood that he won approach a certainty."

Yup.

[bobbyboucher]

15. Florida in 2000 was not completely hand counted, unless

you are talking about the consortium count done months after the election. Many counties just re-submitted their totals for the official "recount" done by the state, which is not a hand count at all.

So if that premise is wrong, what about your others.

[MrUnderhill]

16. Yes... I was refering to the consortium count.

So that "premise" isn't wrong.

Are you questioning their results? This is a media-hired group that categorized every single (of millions) vote into several categories precinct by precinct. By contrast, the exit polling is ALSO a media-hired firm, but with FAR less data to go on.

Why would you doubt the first and worship the second?

[TruthIsAll]

17. OK, now let’s hear your next twelve.

1.Exit polls have historically been accurate to within ONE percent worldwide. Fact.

2.The question is a simple one: who did you vote for? Not who will you vote for?

3.There are 51 states and thousand of counties and precincts. Of course there can be overweighting of a given demographic in a given precinct. Meanwhile, the screen shots showed Kerry leading among women and men in Ohio.

Why doesn’t Mitofsky release the data. Why did he not show at the hearing?

4.I see. Democrats vote early, Republicans vote late? Why don’t they use that excuse in Europe?

5.You are saying that poll workers and/or voters may be lying? Oh. So democrats lie?

6.There is no paper trail in touchscreens. Thanks to the Repubs who control Congress. Why?

7.The Exit Polls were right in 2000 – initially. They had Gore winning Florida easily until Volusia/Diebold dropped 16,022 Gore votes . As soon as that happened, Bush's cousin John Ellis at FOX called it for Bush. Gore won Florida by at least 50,000 votes.

8.Poor Polling? Right. And the poor polling virtually always worked in favor of Bush (41 out of 51 states). Not poor polling. Poor counting.

Now let’s hear your next twelve.

[MrUnderhill]

52. You haven't adequately responded to ONE... why give you 12 more?

1) Exit polls in the US have NOT been that accurate. You continue to spin that the word "exit" somehow makes it irrefutable and immune to error. The polls are done with a different methodology by different groups with a different priority (election results as opposed to issues) yet they are immune from error? If I start doing a poll from my desk under the "exit" sign and therefore call it the "exit poll" will you accept it as inviolate?

2) Yes, the ONE AND ONLY advantage of an exit poll is that it gets rid of the "likely voter" calculation and replaces it with "actual voters"... but it has several other potential error sources - the chief of which being that it is NOT RANDOM. And if that lack of randomness is systematic... you will get an unexpected error in (primarily) one direction. A couple percent shift is very easy to make.

3) You haven't made any coherent argument on this point (and essentially argue as I predicted that the exit polls WERE accurate). Do you or do you NOT believe that women made up 57% of the electorate? I mean.... SURELY these magical exit polls can reliably tell gender to within that same fraction of a percent that they can in GERMANY, THE UKRAINE (etc. etc. etc.) right?

Why doesn't he release them? Who knows?! Zogby doesn't release HIS raw data either. But how can you tout his work product as perfection and accuse him of hiding something in the same breath?

4) I wasn't making the argument. I was pointing out that it is a longstanding conversation re: who votes when. Do older voters vote at a different time? Women? Professionals? Students? I don't happen to know one way or the other... but it would only take a couple percent variation in a couple of those assumptions to throw off the early exit polling. And it's certainly a factor you've ignored in your "model".

5) Nope. I'm saying the pollsters could be lied TO. All I need is the polling methodology and locations and I can throw the numbers off with VERY few people. You really think that there's a guy out there willing to slash the tires on a bunch of vans that were going to pick people up to vote and he ISN'T willing to lie to a pollster? One of those activities is legal and the other one isn't.

All I need to do is drive a van load of people to a few precincts I know will be polled. I know they are stopping every Xth voter coming out. So I send my guys in and signal them when to come out of the door. Five or six responses when they're only polling 100 people at this precinct should do it for me. You see... the pollster ASSUMES that since you came out that door... you MUST have voted (but that isn't necessarily the case). This is oversimplified, but a possible way to cook the polls.

Alternatively, you could just have one or two guys inside the polling firm. That's what the polling firm (who you tout as heaven sent) claims happened. Some people who didn't know how to read the raw data released faulty number to the press.

6) Correct, but they won't be able to hide it anyway. There aren't enough of them (certainly not in Ohio) to keep it from sticking out like a sore thumb. A 300,000 vote theft needs to be spread pretty thin to avoid obvious detection. If the hand count of the paper/punch ballots is substantially correct, you would have to argue that the electronic counties are WAY WAY off (when they have turnout and partisan results very similar to the non-electronic counties).

7) Nope. The early exit polls had Gore winning Florida 3-5%. No reasonable estimate shows Gore winning by anything like 50,000, but even if he did, the exit polls were WAY off. We COUNTED that state. Every SINGLE ballot was categorized and tallied several different ways. Gore may have won by a few, lost by a few, or won by several thousand... but he did NOT win by 3%.

8) Yes... that's exactly what I'm saying. Not that it DID happen, but that it COULD happen and would be a major possibility you ignore. IF their methodology was off on something as simple as the proportion of exurb/urban precincts to choose, it would taint MOST of the numbers ALL in the same direction. That was my point. LOTS of the potential gaps in your "model" would be LIKELY (if they occurred) to impact things in one general direction (NOT screw up randomly 50% each way). I have no idea HOW it could happen, but if the exits really DID think 57% of the voters were female, you would expect that to lean the end results by about 1.5% in Kerry's favor in a pretty uniform fashion.


The early exit polls have been off many times over several elections. And by well outside the MOE. Your presumed argument that ALL of those races were somehow cooked doesn't play well without some evidence.

You also ignore those of us on the ground who know what we're doing. I was active in VA politics for years before moving to NC. We could tell you from the returns of just a few precincts what you were likely to see in the final results. We KNEW where our votes were and how many we could count on. There is actual science to this game. I could tell you right off the bat that the exits were funky when they said VA and NC were "in play" and one was likely to go to Kerry.

[Nederland]

18. Good Post

Although I don't expect TIA to respond to your arguments. I predict a simplistic repetition of "exit polls are always right".

[TruthIsAll]

19. And of course your simplistic avoidance of the facts.

You guys lose the Math argument. You admit its correct.
So you are left with what exactly?

That Germany and Ukraine are wasting their time doing exit polling?
Tell that to Colin Powell.

That Bush beat the odds?
You can't have it both ways.

That the MOE is not accurate?
Tell that to Mitofsky.

That Twenty five years of exit polling experience is useless?
Tell it to Mitofsky.

That no one believes Exit polls?
Tell it to Rove.

You have lost the math argument - that's circumstantial.
You ignore the thousands of documented machine gliches - the vast majority of which favor Bush (86 of 91 touchscreen voted Kerry and came up Bush).

You ignore the 2000 stolen election as if it was a one-time only event.

You are left with nothing.

[Peace Patriot]

26. I would like to know why...

...the TV networks began polluting the pure Exit Poll data with the Republican-controlled electronic official "results" late in the afternoon, thus hiding Kerry's big Exit Polls numbers from the American people? Why didn't they put the two sets of data on screen separately, so that all could see the discrepancy (as they could in the Ukraine)?

And who made this arrangement--to feed the official "results" into the Exit Poll data, and why? (AP, the rightwing Republican owners of the electronic vote tabulators with their secret source code, and the TV networks are implicated.)

This is one of the most disreputable failures of American journalism I've seen--since the leadup to the Iraq war.

[Igel]

32. "pure exit polls" aren't pure.

You don't get useful data from just knowing how many people voted which way. You could go to a bunch of 90% Kerry precincts and say Kerry won with 90% of the vote; you could go to a bunch of 70% Bush precincts and say Bush won with 70% of the vote.

But it's unlikely that you'll randomly pick the right subset of precincts to account for how various groups and subgroups vote. You try, but you'll probably be wrong. Instead, after you get your raw data, you have to adjust it. You need to know turnout. You need to know how to adjust for any differences between 7 am voters, 1 pm voters, and 7 pm voters. You need to know how to adjust for difference in voting patterns between 18-year-olds and 80-year-olds, and how this differs from rural to suburban to urban areas. You'd probably try to adjust for most of this based on historical voting patterns, combined with early election results from the precinct. But past performance doesn't necessarily predict the present.

People refuse to answer. You can't skip them, because if they form a class of respondents you systematically exclude them. So you have "blank" forms. You hope to define that class accurately enough and to get enough respondents in that class to be able to substitute something for those blank forms. The more classes they form, the harder the task.

BTW, exit pollsters sometimes get it wrong. Mitofsky ran an experiment with one NJ elections and miscalled the governor's race by a fairly wide margin. (He used 3 different forms--short, medium, long. Turns out for whatever reason a lot of people refused to complete the short form, and those people just happened not to be a random sample--they preferred one of the candidates. Unreasonable, but that's what the stats force as an assumption.)

[RaulVB]

39. The problem with your facts...

Ukraine's election was contested by Bush's regime BASED IN EXIT POLLS.

Georgia's Shevarnadze WAS MADE TO RESIGN BY BUSH'S REGIME BASED ON THE EXIT POLL RESULTS.

You want more facts?

[Igel]

48. With a very wide spread between

poll results and vote results, not based at the county level but the country level. And with lots of presumably impartial eye-witnesses confirming that things were seriously screwed up, lending credence to the poll/vote difference resulting not from bad polling methodology, but bad polling-place methodologies.

[RaulVB]

50. Oh, oh...

So Ukrainian witnesses tell the truth but "American Democrats are a bunch of liars"?

And the Republicans wouldn't do something like that...

OK, you "win." Game over.

[Igel]

59. Give me an example of

Dems beaten up at polls, acid poured in ballot boxes in Dem areas, eye-witness reports of systematic multiple voting and eye-witness reports of thousands of votes disposed of, etc.

We have comparatively few instances of facts, and those we do have are frequently enigmatic. Votes switched for candidates, but we don't know how or where, but we have guesses. We have truly bizarre absentee counts and registered voter/votes cast numbers; but at least one instance of that turned out to be quite explainable. We have inferences and statistics that I don't know enough about to have much faith in. Why? Because all exit polls have non-random samples; knowing how to adjust for the non-randomness is key.

No matter what I want the result to be when I'm looking at data, I've done too much research not to be well aware of my data's shortcomings and to consider alternative hypotheses, no matter how little I like them or even despise them.

[Nederland]

73. The problem with your facts...

Is that the exit polls used to demonstrate fraud in the Ukraine were FINAL exit polls, not polls posted before half the voters had voted.

[Igel]

29. But they didn't.

They lose the arithematic argument. The math argument doesn't exist. The exit poll numbers are based on assumptions that have to reflect reality for them to be realistic.

I can't defend Mitofsky's assumptions. (I'll try to summarize and post some of his methodology from past polls if I can.) I don't even know his assumptions. I haven't seen them stated anywhere. Some we can make reasonable guesses at.

BTW, my favorite example of exit polls: the 1990 gubernatorial elections. Feinstein had a "healthy lead". The California Poll had predicted a Wilson win. VRS subscribers took the exit polls seriously; it wasn't until 10-11 pm that the exit polls were "weighted", and predicted a narrow Wilson victory. One problem: "exit polls tend to under-represent older voters, who are more likely to refuse to participate, and who tend to be more conservative and more likely to vote for the Republican candidate." The media people fought among themselves--many wanted to call a Feinstein win. They didn't know "VRS does not weight its preliminary results. The consortium will let the subscribers, like KRON-TV, see what these results are, but they are the raw data. And they are not to be taken seriously. Only later will they be weighted, and only then are they useful for predicting the results."

The saner folk prevailed; nobody called the election until later that night.

[RaulVB]

41. For the love of...!!!

THEY DID NOT POLLUTE THE DATA???

How many thousands of posts have demonstrated that they did?

How about some honesty?

[Igel]

46. Arithmetic is fine.

I've solved many an equation and come up with the "correct" answer, only to have it be completely wrong. All you have to do is make a wrong assumption along the way.

The real argument over the exit polls has nothing to do with margins of errors or calculating probabilities, however comforting those might be. The real arguments are over the validity of this particular batch of numbers (the raw data available throughout the afternoon into the evening) and the kinds of adjustments Mitofsky did late in the evening: and all of those boil down to the assumptions that went into where and how to collect the raw data, and what adjustments were necessary.

But I also know that the raw data are pretty pointless. Want me to set up some plausible scenarios that are kasha before you know turnout numbers? (All of which depend on some assumption as to how you'd weight the raw data.)

I've watched all kinds of arguments on DU in the last month. I haven't found them worth discussing in detail because the really important parts of the discussion haven't even been raised.

[Name removed]

47. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Igel]

49. Sorry, dude, accusations of

lying are uncalled for. You can accuse me of ignorance, but I don't see the need for imputing bad faith.

I also don't seem to be able to parse "A foreign power wouldn't have ALL KINDS OF ETHICAL AND POLITICAL ARGUMENTS if demanding Bush's resignation based on the exit poll discrepancies."

[Name removed]

51. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Igel]

60. Ah.

We go from "lying" to "Republican."

I don't "win". This is all an empirical question; I'm not betting. I'm not trying to be Mulder in this debate; I'm trying to be Scully.

[Nederland]

71. Are you incapable of Listening?

You don't have to agree with me TIA, but the least you could do is accurately restate my argument. Its very simple:

The 4pm exit polls were shit. Good math applied to shit still yields shit.

Why do I believe that the 4pm exit polls are inaccurate? Because Mitofsky--the guy that produced the 4pm exit polls--said they were. Do you trust you Mitofsky? It would seem from the above post that you do...

[Name removed]

72. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Nederland]

75. Wow

Stunning the way you ignore my assertion that the exit polls were shit. Care to actually address that little problem?

[RaulVB]

76. No, I don't waste "ammo" in little birds (n/t)

[Nederland]

80. No, its because you HAVE no ammo

You are incapable of proving that the 4pm exit polls are accurate. The reason that you can't prove it is because you are caught in a hopeless trap of logic. Its very simple:

FACT: Mitofsky did the exit polls
FACT: Mitofsky said the 4pm exit polls were shit.

Logically therefore, there are two possibilities:

1) You trust Mitofsky. If so, you have to explain why you believe the 4pm exit polls are accurate even though he says they aren't.

2) You don't trust Mitofsky. If so, you have to explain why you believe in exit polls produced by a man you don't trust.

Its simple and inescapable, but either way, you are wrong.

[RaulVB]

83. Already explained. BTW, to achieve 100% of certainty in a logical...

statement is not possible.

"Either way you're wrong..."

Review your High School notes.

[Nederland]

85. Already Explained?

Nice try. Care to provide a link as to where? Didn't think so.

[Name removed]

21. Deleted message

Message removed by moderator. Click here to review the message board rules.

[bj2110]

24. Real evidence like what? Bush admitting it on Fox?

In all of this mess, with Madsen & Curtis & Fisher, statistics is the single viewpoint that cannot be blamed for biased investiagtions, alterior motives, or bad jurnalism. We have to trust the math. It's the single underlying basis for a case for fraud. I am of the opinion that it most definitely occurred. We're not going to get an admission of guilt from the administration. It obviously won't happen. How about an admission from certain secretaries of state, or certain congressman. Won't happen. What's left that will without a doubt prove that this fraud took place? Statistics. Yes, statistics proven under valid assumptions, but statistics nonetheless. Instead of knocking down what is our only avenue, provide advice on how it could be organizing more accurately. Point out and correct the flawed assumptions, don't just say that they could be flawed. Of course they could be. And if you're waiting for solid evidence, I'm not real sure what else you're waiting on. We're not dealing with isolated bank fraud here, this is fairly large scale, and probably devised by many people who are much more intelligent than you, I, or anyone else who posts on this board. We have enough to go on right now. Hence the Arnebeck lawsuit. Support that. Support the underlying statistics.

[Name removed]

27. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Name removed]

57. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Nederland]

84. Here's a CLASSIC example of TIA getting it wrong

Mark this thread. I said it before. I will say it again. Bush won't run.

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=104&topic_id=1008784


Yup, TIA sure predicted that one correctly...

[Imajika]

89. That is indeed a good example...

...of TIA getting it wrong, but then that is pretty much par for the course.

I can't actually think of anything TIA has predicted that has come to pass. As best I can tell, the guy is wrong about virtually everything.

The fact that so many people here actually buy TIA's "research" is simply mindboggling. Trillions to one odds that Bush really won? Come on, it's utterly laughable. Do people really think that with the odds trillions to one that Bush really won, that no one at a real polling organization, the DNC, the FBI, the MSN wouldn't be on to it themselves? There are some some pretty clever election experts working for the Democratic Party and Kerry campaign. These professionals are not stupid people, infact they are the best of the best - and they told Kerry that the votes simply were not there for him to win Ohio and the Presidency. All the blather here and on other weblogs claiming that the provisional and absentee ballots would deliver Kerry huge numbers of votes, putting into play the under/over/spoiled ballots was completely wrong - just as the Kerry campaign people understood.

I am not saying the Republicans couldn't have cheated on some massive scale. I suppose, though seriously doubt, that this could have happened. But to believe that the GOP stole election 2004, real evidence of this would have to be presented - not bizarre conspiracy theories with zippo to back them up.

I wonder when TIA will start predicting that "Bush is toast" again? Heh, at least he can safely predict that Bush won't run next time around - he would at least get something right.

Imajika

[Name removed]

97. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Wordie]

25. I gave you one possible cause last night: circumstances in this election

were vastly different than for any of the earlier elections on which the pollsters' models were based. For instance: what was the effect of ObL's tape, or 9/11. How would it even be possible for the pollsters to work into their models something to account for the possible effect on this election of these things, since there has been no previous election in which things of this magnitude could have been a factor?

In other words, the circumstances of the 2004 election were just too different for the pollsters' models to account for. So even if every previous exit poll had been right from the beginning of time, they could easily have gotten it WAY wrong this time.

BUT, I want to make clear, I am presenting this as a possibility, not a fact. You had asked earlier for other possible explanations of the discrepancy between the exit polls and the results, so this is one possible explanation. That said, I must also say that I just DO NOT KNOW the truth of this. There may be something to the exit poll theory or not.

My problem is that we can't discuss this if every time someone posits an alternative explanation they are "called out" for not supporting the fraud stories. I believe that one of the best ways to arrive at the truth is through honest discussion. To my knowledge, there is no litmus test regarding such beliefs here on DU, and there should not be. But I am increasingly feeling that you are trying to create one.

[RaulVB]

45. I know is not directed to me...

Sorry to interject.

Circumstantial elements of any election SHOULD NOT REFLECT IN SUCH A DRAMATIC WAY THAT STATISTICAL PATTERNS ARE INVALID, therefore voiding the WHOLE MODEL APPLIED TO THE STUDY OF VOTING RESULTS, same model that was accepted as "legitimate" throughout American history.

You need to understand that is just impossible to think that way.

Your other points are mute after the basic facts that I just described for you.

[Wordie]

55. But Raul, you don't have any way to prove/disprove your point. It is an

opinion, not a fact. How do you KNOW that the circumstantial elements in this particular election should not be reflected in the poll results? And I didn't say it voided the whole model, just that it may have been enough of an intervening variable to account for the results. Its not as if either candidate, no matter how you view this, won in a landslide. The results were close.

And its not impossible to think that way, because I *could* think that way, but even with this possible hypothesis that I have presented, I would want more facts. I would want to *test* my hypothesis, to expose it to scrutiny to see if it may be true.

[RaulVB]

58. I see you point, believe me...

Your argument is reduced to this:

"Have some faith and trust the people running the process..."

You do. I don't.

I can't prove that I'm 100% right. You can't prove that I'm wrong.

We agree on that.

[Wordie]

64. You've made a HUGE leap, from my wanting to test hypotheses, before

presenting them as facts, to my trusting the people running the process! You have stated as a fact, something about ME that just simply is not true! That is exactly the same sort of what worries me about TIA's analysis. I am afraid he is making those sorts of huge leaps too.

Here is my point: there is so much that it appears to me is PROVABLE, why muddy up the waters by combining it with purely speculative material? Why weaken a strong case by adding weak arguments in support of it?

[RaulVB]

65. And exactly there is where I disagree with you!

I don't think is valid to focus in a "methodological controversy," at this point.

I would call that "muddy up the waters..."

[Wordie]

66. And what if one of the people at these hearings were to believe TIA, and

everyone was to hear about it, but then, because the methodology was WRONG, TIA was discredited, even by people on our side. Those other people, the ones we need on our side, would be discredited too, wouldn't they? What then? The entire issue of the voting problems could go down the drain if one of the major players in the thing trusts one of these speculative arguments that can be proven wrong. Is that what you want??????

Statisticians test their hypotheses with a whole variety of sophisticated tests that have been designed for just such a purpose. TIA, to my knowledge, has not done that.

VERY intelligent people, as TIA clearly is, frequently misunderstand statistics. It is very easy to see a correlation and presume a cause and effect that just isn't there. THAT is what I am worried about. I don't want anything undercutting the investigations that are going on.

[RaulVB]

74. We have to agree on the basics...

And I think we do.

I don't think TIA pretends to be the "owner" of the evidence. His efforts are hugely significant but they don't exclude the works of others, and we all understand that.

Fraud took place and we have ten thousand tons of arguments and facts to illustrate it.

[Wordie]

78. All I'm saying is use ONLY those arguments that can be verified, and not

those that are speculative. That's all. Why risk the tons of verifiable material, by using the unverified stuff? Because you know what the other side will do, doncha? They will seize on ANY mistake and you will never hear anything else. The 99% that is verifiable will never see the light of day, because they will be so busy - screaming and yelling and playing it over and over (like the Dean scream) - using that 1% that we put in there that couldn't be verified to refute the ENTIRE thing!

[TruthIsAll]

98. OBL's tape makes this election different? Very weak argument.

One can always say an election is different. All elections are different. But, exit polls are always consistent - across all elections and all continents.

The mathematics of probability is always consistent and has been since Bernoulli and Gauss first developed the theory - and the pit bosses in Las Vegas know it. No one, I repeat no one, disputes the mathematics.

The only thing left for the ditto-heads is to throw out the strawman that the exit polls themselves were faulty. Then I will just ask them to explain the 86 of 91 touchscreens that turned Kerry votes to Bush.

The mathematical probabilties along with the DOCUMENTED evidence of voting machine anomalies, ALL IN BUSH'S FAVOR, lays out a very solid case alone. And that's not even including the voter registration fraud, the spoiled punched cards and the fact that the vote counting was done by Repukes committed to Bush.



That's a lot more credible than your specious argument that the election was "different" from other elections. It was only "different" in the size and scope of the fraud perpetrated against the American people.

THAT IS WHAT YOU SHOULD BE VENTING ABOUT.

[jkd]

101. Exit poll accuracy?

I made this post before on another thread, but no body chose to answer. I'll try again?

Concerning BYU’s exit poll accuracy, Dr. Freeman said this:

“True to their word, predictions in this year’s contests were quite accurate. In the Utah presidential election, for example, they predicted Bush 70.8%, and Kerry 26.5%. The actual was Bush 71.1%, Kerry 26.4%. Consistently accurate exit poll predictions from student volunteers, including in this presidential election, gives us good reason to presume valid data from the world’s most professional exit poll enterprise.”

Mitofsky had Kerry at 30.5%. Isn’t Dr. Feeman contradicting his own argument?

[MikeDuffy]

5. "one out of 256" or "one out of 1024"?

In Freeman's follow up paper of Nov 21 (http://www.buzzflash.com/alerts/04/11/Expldiscrpv00oPt1.pdf)
on page 9, he calculates the probability of tossing heads 10 times in a row. Should that not be (1/2)^10 (one out of 1024)? (I am not qualified to comment on his statistical analysis.)

[TruthIsAll]

7. (1/2)^10 =1/1024 = 1/2* 1/2* 1/2* 1/2* 1/2* 1/2* 1/2* 1/2* 1/2* 1/2

You are right.

It's 1 out of 1024.

[electropop]

10. I was puzzled by that myself. n/t

[americanwoman]

102. Freeman corrected coin toss in latest version...

at http://www.appliedresearch.us/sf/Documents/ExitPoll.pdf

[Name removed]

6. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Helga Scow Stern]

9. What do you have to sat to Kim Zetter's article from 12/7?

http://www.wired.com/news/feedback/mail/1,2330,,00.html

Just so I can have some ammo.

[TruthIsAll]

20. Sorry. I'm not familiar with it.

.

[Nederland]

14. Its not the math that is wrong

Its the numbers that you plug into the equations. You have always ignored this fact and I expect that you will continue to do so.

[AnarchoFreeThinker]

22. can I open another can

of worms? What are the odds that both the election day exit polls AND the last PRE-election Zogby polls were BOTH not indicative of what actually happened in the key swingstates?

[Jimr]

23. I hope you're right but...

HI
I hope you're right.
I think the weakness in your argument might be the assumption that the discrepancies are normally distributed.
It has been many years since I studied and used statistics but if I remember correctly; a measure that is normally distributed usually has a large number of additive factors. Predicted vote count based on stratified samples used in the calculation of the discrepancies can be easily skewed by a few mis chosen factors.
You would probably have a stronger case using a nonparametric test like the Wilcoxon Signed-Rank test.
Jim

[TruthIsAll]

31. With all due respect, why don't you just do it?

I would love to see your analysis.

[Jimr]

35. Time

As I said I'm not a statistician.
I'll do my best and look in to it on Monday.
Jim

PS Thank you for your efforts.

[TruthIsAll]

103. You might be interested in these Time Zone stats

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=203x141048

[Name removed]

30. Deleted message

Message removed by moderator. Click here to review the message board rules.

[Mastiff]

42. I have to call you on this

Exit polls are normally very accurate. Will you accept this as a given? They are sampling those who voted right after they did so. These are used all over the world to call elections. They have also been used in America, accurately predicting results for at least a generation. Why would Mitofsky set up an inaccurate exit poll? This is what you are claiming he did. However, the exit polls matched the outcomes in many non-swing states. The swing states are the ones that are not matching. You can't claim it was some methodology on the pollsters part, as if it were the results would be a uniform bias across the board, not just in the swing states.
I personally don't care if the odds against the polls being off as far as they are is 1 in 13T or 1 in 10,000. The result in any case is highly improbable to next to impossible. What is important is that these results raise questions that the tabulated results are in error. This administration has used exit poll data to call into question the Ukraine election. Why can I not use the exit poll data to call this one into question? The results are questionable. We know that the GEMS tabulators are subject to hacking. We know that 80% of the US vote is counted by two firms, owned by two brothers, who just happen to be strong Bush supporters. We know that there is no paper trail on many of these votes, so a recount is impossible. We know that the chief elections officials in Florida and Ohio are strongly partisan Bush supporters. We know that votes were suppressed in minority precincts in Ohio. We have many edicts from Blackwell intended to diminish both the counting of provisional ballots and to disenfranchise new voters.
When I went to bed at 1AM on election night I saw the CNN graphic with Kerry leading in the exit poll in Ohio. When I get up the next morning he now shows as having lost in the exit polls. This was not possible given the difference in the number of respondents. This gives rise to a feeling that something is not right here.
You can question the math all you want to. The facts are not possible to dispute. You may spin as much as you want (which, in my personal opinion, is all that your comments amount to). We have relied on exit polls for years to be accurate. There is too big a discrepancy between the swing state differences to the non-swing state congruence to automatically assume the bias is because of bad exit polling. The opportunity to hack this election was there. I think the possibility that fraud existed is much higher than that the exit polls were uniformly bad. For you to say it was bad exit polling and denigrate the possibility of fraud does not seem to me to be at all honest.
Have a nice day.

[stella2cat]

61. very well said, Mastiff! thanks n/t

[Name removed]

96. Deleted message

Message removed by moderator. Click here to review the message board rules.

[platinumman]

100. Why did you bother to ask?

I have no idea why you asked for the comments of mathematicians if you only intended to insult their response.

As has been said many times on this thread, your maths may be right, but your assumptions are not, particularly your assumptions of randomness, which make your maths irrelevant.

Any analysis from the figures that are available from the exit polls are meaningless if we do not know the design of the poll. Many people, understandably, have compared the US exit polls with those of the Ukraine. However, the Ukraine polls were designed to check the integrity of the election. I think there should have been such a poll in the USA, but there wasn't. The US Polls were not that at all, but a commercial undertaking to provide a degree of prediction to the media, post-election analysis, and, of course, profit to the pollsters. And they had to meet these aims within their budget.

Let's have a look at your own exit poll. One that I am sure you carried out. Let's say that among your acquaintances, you would expect 20% to vote for Bush, or maybe 20% voted for Bush last time round. (yes, I doubt if you even know any Bush voters, but never mind). Now, if you asked your acquaintances on the evening of the 2nd who they voted for, and you found out that 30% had voted for Bush, you should start worrying that Bush had won the entire election, even though, in your own little poll, Kerry was winning easily.

That is more or less how the Mitofsky poll might have worked. (I say 'might', as we don't know, but it is how such polls are conducted in the UK, and it is a straight, possible, explanation to the figures that we have). They are not going to send pollsters out into the countryside, as they would be standing around doing nothing most of the time. Instead it is far more efficient to keep them in the urban areas of most states. They know that these are mainly democratic voters, and the figures obtained would reflect that fact. However, the important figure is the 'swing': in other words, how much the measured lead of the Democrats is greater or less than the lead which would be expected in those areas to win the election. The measured lead may be 3%, but if one would expect 5% in that area, then that 3% would mean a prediction of a Republican win. However, it would take time to work out, and margin of error is necessarily far larger than that which would be obtained by pure one-stage random sampling, but it is enough for the temporary needs of that day. (If you think that such a predictive poll is a waste of time, I would not disagree with you: it serves commercial purposes only)

Therefore, as I said, any analysis without knowing the design is spurious. You may well ask why Mitofsky does not produce the design. It may well be that when/if he does, the figures will 'indicate' some fraud (it could never be 'proof'). It may be that they screwed it all up and he is professionally embarrassed. There is a lot of competition in the polling field, so it may be that he is keeping it secret as a matter of safe principle. Of course, if there has been fraud (and I have to think there was) one would hope that it does show up in the exit polls, but one cannot be sure of that, as the swing necessary to return the election to Kerry would probably be less than the margin of error.

However, there is a simple explanation for the odd figures, which they can trot out at any time and I fear that if you constantly bring up this red herring, then, in the eye of the public, the stronger claims of fraud proof will be damaged. Such important claims are the lack of voting facilities and the clear opportunities to corrupt the counting software. Given time, such claims are verifiable, through inspection of documents, regression analysis, and, over time, auditing and compulsory inspection of source codes. I'm sure all these things will happen, but it is a waste of time, for now at least, to try to claim proof from the exit polls.

[BamaBecky]

99. My head was swiming till I read your post.....thank you & TIA n/t

[shaggy briard]

53. Not a math question

But a competence question. If exit polls were a pure random sample of all voters then all of the above math would be right. However, exit polls are not conducted at all precincts. Instead, the polls are conducted among a non-randomly selected set of voters at selected sets of precincts presumed to be representative of the electorate as a whole. Therefore, there is the possibility that the sample was consistently screwed up in identical ways across many states because fundamental assumptions that went into the selection process were not borne out in the actual voting patterns --
Example: If African-American precincts had consistently low turnout, they would be weighted for low turnout going in and under-selected in the samples. If, however, they had exceptionally high turnout, then the weightings would be off and actual results would consistently be off in favor of Repubs in a large number of states. This is pretty much what happened in the 1990 and 1992 mid-term senate election Similarly: If rural, Amish counties were traditionally under-voting but showed up in large numebrs this election, then they would be under-represented in the exit polls and their impact would be under-estimated. This may have happened in Ohio, PA and FL in this election.
So, either the exit pollsters need to admit that they made a mistake in selecting and weighting their "sample" precincts or the exit polls were right. I personally believe that it is a little of both.
a) Turnout was not as expected -- therefore samples off
b) Fraud on a significant scale was conducted and polls were essentially correct
Both can be simultaneously true.

[papau]

62. conducted among a non-randomly selected set of voters ? - every 10th

voter was asked in the selected voting locations - I do not see a bias.

If you are saying the selected voting locations had a bias in there selection, you know more than I do since what do know about the selection has no bias.

I grant you the their was an implication from an early interview with the 04 exit poller that he found the Florida 00 voting location selections to be suspect - which may imply that he may have run some cross tabs to get "good" polling locations - as opposed to a random method - I just do not know.

As to "Turnout was not as expected -- therefore samples off" - I have yet to see any data that says the turnout was different than the model. I do see that proposed for non-obvious characteristics
such as race and sex - but on the obvious items - the deviations would seem to favor a greater Kerry vote.

So I think we come down on "Fraud on a significant scale was conducted and polls were essentially correct"

I reject as silly, based on the stats, the excuse that GOP folks are shy (there was no global effect) or that location chosen by exit poller always favored a Kerry response (huh - if any know how to chose such a location, please post!).

:-)

But indeed -
Both can be simultaneously true

[Mastiff]

79. Just a question.

I just looked at the Colorado exit poll results. There was a 3.8 swing from Kerry to Bush. However, the Senate exit poll matched perfectly. Saying there was a bias means the Senate race should have deviated as much as the Presidential numbers. In any case it would have not been a perfect match. If there was a bias how could you have matched the Senate race perfectly? Likewise, saying it was a fair poll on the Senate race calls the Presidential figures into question.


http://www.exitpollz.org

[papau]

87. very true - nice catch :-)

:-)

[shaggy briard]

91. Non-randomly

In this case means that they chose where to send the interviewers based on some criteria -- usually based on prior turnout -- since they didn't send interviewers to every precinct. At the polling stations, the polls are conducted randomly. Question comes down to how well the pollsters chose the precincts to sample. So, for example, if turnout was up 20% in Kerry-leaning precincts and 30% in Bush-leaning precincts, there could very well be an undersample of Bush voters.
I do not believe that people lied or refused to answer either -- except maybe in New Hampshire -- where there are anecdotal stories of people doing that before.
I suspect that the exit polls were substantially accurate -- however, the way it is framed as a "statistical likelihood" of all being wrong in the same direction presumes a baseline that doesn't exist. Let's face it, the bastards stole the election.

[Wordie]

86. Very good points you make here. I think there is a basic misunderstanding

about how the sampling is done for these polls, and that may be part of the problem.

[brindis_desala]

56. TIA was supported by the Economist

which showed Kerry consistently ahead from early October.
http://www.economist.com/media/pdf/yougovS.pdf
I submit that exit polling alone in the ABSENCE of any supporting evidence would not automatically constitute fraud. But TIA point is impossible to argue: given that the polling was overwhelming skewed in one direction; that the poll adjustments made on television were impossible given the small increase in sample; that the exit polls cohered with the actual result in most non-swing states and senate races; that voters REPORT trying to vote for Kerry and seeing their votes transferred to Bush; that so called "glitches" ALWAYS occur in Bush's favor and in rational numbers; that machines routinely broke down and were attended by poll workers and technicians; that the machines themselves are made and owned by Bush supporters; that the Governor of Florida is Bush's brother and the SOS of OHIO his chief campaigner; that there was rampant voter suppression in Democratic swing state counties; that the lines were longest in Democratic precincts suggesting a late swing for Kerry and finally the unprecedented GOTV campaign by grassroots Democrats. Unless circumstantial evidence is no longer a valid legal argument- AS far as this argument goes: TIA wins in a landslide.

[alittlelark]

68. The Economist is one of the few magazines I read every week

even 'The Nation' goes unread occasionally. The Economist is the most respected mag I know of.

I think we might need to consider blitzing them w/ these #'s... they have been brave enough to stand up to, and against b*shco for about a year now.

[Wordie]

81. Not a bad idea. They probably have statisticians on staff who could

evaluate the analysis done by TIA, or at least they may know of one who they could enlist. Statistics is very tricky, and there is a lot involved that is counterintuitive. So that would be helpful.

[emcguffie]

82. there is one more thing --

And that is the pattern of anomalies, not just with the exit polls, but with what might be expected, because of, say, voting patterns, and what happened.

Or, trying to vote for one candidate, and getting the other. I honestly didn't hear the complaint once on election day of anyone trying to vote for Bush and getting Kerry. And they would have complained very loudly. Very loudly.

I don't have at my fingertips all the analyses that people have done. But when you put them together, it is kind of convincing.

Oh, and the exit polls could have been wrong because Mr. Mitofsky wanted them to show Kerry winning. But why would he do that? Just to make us all go crazy?

But, gee, I think I've also been called a troll. Although I'm not sure. I do believe there was widespread fraud. I believe there was fraud in 2000, with more in 2002, and more in 2004. And we have to stop it now or we'll never have a legit election again. And we will have a one party state. Soon.

[platinumman]

88. Heads on a hard wall

I'm not part of the US political scene. I'm British and I live in Asia. However, the results of this election affect us all. I should say from the start that I find the US election system absurd from beginning to end, inside and out. And unless it can be verified that there is no fraud , I can see no reason for assuming that there wasn't. However I get disturbed when I see people insisting on hitting their heads on a very hard wall. The exit polls may well be able to indicate that there is fraud, but as we know little, or even next to nothing about how these figures were derived, there is little analysis that can be done. The assumptions that have been made by some on this list could well be used against you by those wishing to hide the real truth.

We have no idea as yet what sampling was done on the day, and there are many possible, legitimate scenarios which could have led to the early, or even the late, unadjusted figures. Most obviously, the pollsters could have taken a uniform ratio of urban to non-urban areas in every state, which could have led to the same error in many states. (This has been already mentioned in this thread). This sounds dumb, and if the polling company had the single remit of verifying the presidential election, it would have been very dumb. However, in this case, there were probably many aims and issues to be covered; there were many elections to check; and there was a limited budget with which to carry out these tasks. They had to make a profit from these polls, and the profit lies in getting a lot of responses about a whole pile of issues which destract from the role of verification.

There has been comparison with the exit polls in the Ukraine. However, those polls were designed as single issue polls, purely for verification purposes. And we know how they were carried out (or at least it can be found out). Of course Mitofsky should be further pressured to make the data and methods known, but I would suggest that one reason that he is not divulging is that he doesn't want to be professionally embarrassed.

I'm sure that the exit polls might eventually produce some valuable messages. It would be revealing to know, for example, if there is any significant differences between responses for state elections and federal elections, and of course when you do know what the sampling methods were, you can check results with the actual precincts.

But until then, I think you would be wise to concentrate on the issues of availability of machines and the anomalies in the actual results. And look for more programmers willing to spill the beans. And get that source code. And then, please, change the whole bloody system.

[donkeyotay]

90. Well said. To my mind, it's the circumstantial evidence

that people work hard to ignore. Bush had the means, motive and opportunity. If they wanted us to have confidence in his alleged election, they should have made it open and transparent. It's the little things like the ownership of the proprietary software and the Bush/Cheney team running the elections, not to mention the GOP media, that makes people doubt a result that was, at best, a surprise.

Anybody remember that the bastard lied his way into a war that may yet destroy our country? Impeachment would have made more sense than re-election. Even the troops and many republicans are pissed at his incompetence, so it's hard to believe that this type of support materialized out of the thin air of a supposedly moral agenda.