A scary item (is it true?) submitted to Dave Farber's Interesting People mailing list by one Ken Deifik:
It occurs to me that one of the questions that could be answered without too much trouble, at least for someone with lots of access to data and a knowlege of statistics, would be: is there any difference in the Bush – Kerry percentages in precincts that used eVoting, especially Diebold but all eVoting machines, as opposed to those that used paper ballots or some other method of voting. If this question has any meaning for you, I'd ask you to pose to the list, to see anyone with the proper skills and access could carry out such a study.
I hope in the next few days statisticians examine the issue of the exit polls. Since the early 70's the exit polls have always been spot on. I feel ashamed for any journalist who says the exit polls got it wrong in FLA in 2000, because it is clear they got it right.
I have to wonder how John Zogby, who predicted 312 electoral votes for Kerry at 5PM EST 11/2, could have gotten the exit polls so wrong. Or really if he did.
One reason may be who votes at what time of day?
I just found this posting in the Democratic Underground site
…on several swing states, and EVERY STATE that has EVoting but no paper trails has an unexplained advantage for Bush of around +5% when comparing exit polls to actual results.
In EVERY STATE that has paper audit trails on their EVoting, the exit poll results match the actual results reported within the margin of error.
So we have MATCHING RESULTS for exit polls vs. voting with audits
A 5% unexplained advantage for Bush without audits.
Say it ain't so…
More disturbing info, if confirmed, from daily kos diary at
that analyzes the difference between 2000 and 2004 in florida, concluding that the numbers just don’t make sense:
Bush 2,912,790 Bush 3,836,216
Gore 2,912,253 Kerry 3,459,293
Nader 97,421 Nader 32,035
Other 40,193 Other 28,382
= 7,355,296 (2004) – 5,963,657 (2000) = 1,392,639 new voters (99% precincts counted, no provisionals or absentees).
So, we have 1.39 million new voters, and Kerry loses by 376,923 votes? Thus, he lost an overwhelming majoirty of them, or he lost an overwhelming majority of regular voters – much, much more than Gore lost.
We have 77,197 fewer third party votes, but Kerry loses the vast majority of these?
Exit polling numbers show that Kerry had more Hispanic and Cuban support than Gore did, and Kerry lost?
Most exit polls in Florida showed Kerry leading, yet he loses by a massive 5%?
I remember Dick Cheney, in an interview, prognosticated that BC ’04 would win 52-47. A five-point spread.
This should be looked into. It probably won’t do those of us who didn’t want Bush around for the next four years any good right now, but the best thing we can do for the future is–by whatever means–strive to weaken the structure where parties with a vested interest determine the procedures for the elections. Even Republicans might appreciate this–after all, they may not be top dog forever, and whoever’s in charge can easily turn the tables with the existing structure.
Look, this time the data are widely available, so a detailed analysis is possible and should be done. In 2000 the VNS forbade their clients from publishing the data OR the study they did that “explained” the Florida errors. [We now know that their exit poll projection of 3% for Gore would have been correct were it not for the mass spoilage of ballots in Duval and the unexpected emergence of the “Jews for Buchanan” faction in Palm Beach.]
And in 2002 VNS just shut down the whole operation without letting anyone see anything.
Now, keep in mind that exit polls are much more reliable than pre-election polls for a lot of reasons (mysterypollster.com goes into this topic in detail). International election observers use exit polls as part of their validation methods. And, as stated above, the 2000 exit poll error in Florida turned out to be explainable as a combination of voting malfunctions. In 2002, lots of us suspect the exit poll data may have been revealing voting irregularities in a number of races that experienced massive swings from the final pre-election poll, but without the data we don’t know.
Of course, even if you can prove massive voting “irregularities” the media and “elected” officials will ignore it. So clearly, the lesson is that in the US the challenger must win by a large margin instead of 2-3 percent.
I had the impression that the page Zogby posted late Tuesday afternoon was NOT based on exit polls, but was rather the result of his final pre-election polling.
I distinctly recall the chart stating that Florida was a couple of tenths of a percentage in favor of Kerry, and “Trending Kerry” — which is something you’d discuss about the latest in a series of pre-election polls, but not about an exit poll, nor would you compare an unweighted exit poll against weighted pre-election polling. Also, by 5 PM, none of the exit polls had been weighted for turnout, making them pretty useless for detailed prognostication (no [lie] Sherlock), and I can’t imagine that Zogby would have staked his reputation on mid-day exit polls. Of course, the page is now replaced with Zogby’s own mea culpa, so I can’t prove a darned thing now.
[Found before I hit the Post button: Zogby is confirming that his posting was pre-election polls. Statement from John Zogby on 2004 Presidential Election Results: We feel strongly that our pre-election polls were accurate on virtually every state. Our predictions on many of the key battleground states like Ohio and Florida were within the margin of error. I thought we captured a trend, but apparently that result didnt materialize. We always saw a close race, and a close race is what weve got. Ive called this the Armageddon Election for some timea closely-divided electorate with high partisan intensity on each side.” http://www.zogby.com]
While on the subject of Zogby, I recall hearing him over the last weekend on NPR, and he was discussing a late poll where he stated that his poll on the (national!) Hispanic vote had settled at 63% Kerry, and that this was over a magic number that people had determined would get Kerry over a serious hump. As I looked at the CNN exit polls from late last night, I saw they pegged Kerry at 56% (nationally) for Hispanics. Presuming that Zogby’s and CNN’s definitions of ‘Hispanic’ are equal (and all of the other statistical and poll-taking anomolies which we all know are out there) — What happened? Did the Cuban vote come in drastically larger than had been anticipated (and, assuming the conventional wisdom of Cubans vote for Bush is true, leads to the total lack of help from the southern FLA counties)? Was that the secret weapon in Florida?
Okay, I understand that you liberals are still licking your wounds so i’ll go easy. Conservatives distrust media. Media conducts polls to damage us. So we just didn’t answer the pollsters when they approached us. So when the exit polls came out they where inflated fior democrats. If you look at the exit polls they got it wrong because of bad math. That’s all. You are right though on one thing though. There is a conspriracy against you, and there are 58million voters in it.
Let’s have a little giggle-relief while deciding whether to don the t-hats:
During his 1956 presidential campaign, a woman called out to Adlai E. Stevenson “Senator, you have the vote of every thinking person!”
Stevenson called back “That’s not enough, madam, we need a majority!”
(Thx to Ceteris-Paribus).
Nigel, I see that conservatives are also very clever. They clearly answer the exit pollers in precincts where there is a paper trail, but refuse to answer in precinct where there is not a paper trail.
Nigel, I see that conservatives are also very clever. They clearly answer the exit pollers in precincts where there is a paper trail, but refuse to answer in precincts where there is not a paper trail.
IT AIN’T SO.
There is another, much more credible explanation for the dispararity between the exit polls and the final vote:.
THE EXIT POLLS WERE PLANTED.
There is abundant evidence that afternoon exit polling favoring the challenger motivates extraordinary late-afternoon turnout from the incumbent’s party. The Ohio Secretary of State was no doubt informed of this trend, which became a factor in his decision to keep the polls open until 10 PM.
See http://www.battlebunny.org/archives/2004/11/why_were_the_ex.php for the gist of the argument.
Apologies – in previous post, that’s “disparity,” not “dispararity” — I’ve been awake for far too many hours, at my advanced age, and truth to tell, I’m emotionally devastated by the election results. But I must say I rather like this misspelling…
I agree with the writer who said she was too old to stay up these hours.
I am too drained also. We found it interesting that we have NEVER received the absentee ballots we had requested well in time for the election.
We voted early anyway – and never received the absentee ballot which should have been mail.ed out long before we voted.
I fear for our democracy. I wonder if the democratic party will survive this — if we will survive this.
Aren’t exit polls weighted to match the sample with an estimate of the voting population? We ran behind the exit polls in places like Iowa and New Mexico, with Democratic governors and no history of cheating. (We even ran behind in states we won like WI.) If only OH and FL were out of whack I’d be more intrigued by the conspiracy theories. Instead, I think the voters who showed up, even as implied by the end-of-day exit polls, were not quite the cohort predicted: more male, and the GOP seems to be leading/even in Party ID among actual voters, which is an increase for them. So the weights were off, everywhere.
But if someone can generate other reasons, hey, I’m all eyes.
NPR reported yesterday that the exit polls seemed to line up a lot better with the punch card counties than the touchscreen ones…not sure what that means, especially if you leave the aluminium millinery at home.
My tinfoil hat’s out for cleaning so I’ll have to say I can’t buy this one. There was just too big a swing toward Kerry in the late pre-election polling in Florida to be really credible and, as Michael pointed out above, the Zogby story was about pre-election polling rather than exit polling and there wasn’t time to have put together accurately normed exit polling data by 5 PM anyway. Of course, the mere fact that it is impossible to disprove this claim, however fallacious, supports adding a paper trail to electronic voting so such claims could be debunked with real data.
This seems silly to me. I’m all for a paper trail for Diebold machines, but the company couldn’t keep it’s crappy security code secret. How could they accomplish a secret conspiracy to swing a national election?
Diebold markets those things through tight connections with the GOP. No surprise, they’ve been adopted more places where the political infrastructure is dominated by the republicans. This is like an ommitted variable in that correlation. States with GOP infrastructure, turned out more conservative voters than expected. End of story.
Well, whatever happened it’s important to find out and understand it. Let’s look at some possible explanations, all of which should be testable via post-election investigation:
1) Nigel’s explanation is possible, because there is precident in elections between a black/white candidates in the South. Some southerners have learned to pretend to support the black candidate when talking to strangers while actually supporting the white one. Pew did a study on this, and of course they found that the people who did this held extremely negative views of African-americans. Pew didn’t come out and say it, but I think the technical term for them would be “closet racist”.
However, I’m very skeptical for 2004 since most of the exit poll errors seem to be concentrated in areas with strong Bush support. That is, while I can see that a San Francisco resident might be reluctant to admit to voting for Bush, I doubt a resident of Tallahassee would feel the same.
2) Pinky’s late-surge theory has also been mentioned by many, with lots of anecdotal data. I’m very skeptical. What this theory suggests is that something like 5-6 million registered Bush supporters weren’t bothering to vote until they got a mid-day call pleading that they vote because Kerry would win.
3) Greg Palast has loads of documentation on the supression of vote counting. There are two parts to this. a) People are turned away by long lines or told they can’t vote. According to Greg this may number in the millions, but none of those should count in the exit polls, since those people didn’t vote. b) People voted but their votes weren’t counted. Again, according to Greg this may number in the millions. Like with Palm Beach and Duval county in 2000, where phenomenally high ballot spoilage and wrong votes, respectively, caused a 2-3-percent distortion in the final vote that explained the error in the exit polls. This seems a highly likely contributor to part or most of the exit poll discrepancy.
4) Vote Fraud. The whole system of patchwork voting methods, inconsistently applied, is rife for fraud. Statistical analysis should be able to detect such frauds by comparing exits to final results on a county or precinct basis and analysing errors, then removing any noise from (3) above.
5) Exit poll error. Early exit poll data should be tossed since the MOE is so wide, except to possibly detect hour-to-hour trends as per (2) above. But the sub final exit poll data (that is, the results BEFORE they are normalized to match the polling results) is what should be used as the basis for analysis. Collection methodology, including how the polls addressed issues of absentee, early votes, and provisional votes, could be examined to see if the methodology introduced sampling errors. Exit polls used to be very reliable but the increased use of alternate voting methods has made it harder to get them right.
See my previous post re: Observer’s anti-southern nonsense, please. To me, there’s very little difference, at least in terms of ignorance, between a racist perception that an entire race’s thought process and ideals can be summed up without a nod to individuality, and a yankee perception that the South swings as one group, evidently however we were told to at the last Klan meeting.
Cubbyholing any person into a group mentality and ignoring their individual ability to think is patently unamerican. Don’t believe me? Read Brown v. Board. ‘Bout time someone did, apparently.