Some interesting poll data, and the lesson thereof

[unixronin]

It was previously reported (at www.electoral-vote.com, among possibly other places) that while actual ballot results matched exit polls to within the margin of error in counties that used paper ballots or electronic ballots with a paper audit trail, counties that used electronic voting with no printed audit trail consistently showed an unexplained surplus of pro-Bush votes in the actual ballot totals averaging about 5% relative to the exit polls.  Now someone has charted ten paper-ballot states vs. eight paperless electronic ballot states, and the results are enough to arouse curiosity.

Unfortunately, this being in PDF form, I can't easily insert the charts themselves here.  You'll have to look at the charts yourself.  Also, note that the charts cover only 18 of 50 states.  I don't know the basis on which these 18 states were selected; per Benjamin Disraeli, one can lie convincingly by cherry-picking one's data carefully.  A reasonable assumption would be that the ten and the eight were selected to most clearly show the result being asserted; however, any conclusion drawn from incomplete data must necessarily be suspect.

However, that said, if verified, and if borne out by the data from the remaining 32 states, the comparison is highly suggestive.  It implies that discrepancies in electronic balloting had dramatic effects upon the election outcome in several states.  It is already known that errors occurred; USA Today has reported, for example, on one Ohio county in which a malfunctioning electronic voting machine generated 3,893 extra votes for George Bush from a single precinct in which only 638 votes were cast, and over 4,500 votes were lost in one North Carolina county when voting officials didn't realize they had overflowed the voting machine's memory capacity.

Now, this chart points out that actual results varied from exit poll results in the eight paperless-ballot states charted by as much as 16%. Notably, Kerry's margin of +17 in the New Hampshire exit polls turned into +1 in the final results, and Bush's margin of +4 in North Carolina exit polls turned into +13 in the machine tally.  More significantly, in two states, this discrepancy flipped the outcome of the election, from +1 Kerry to +5 Bush in Florida, and from +2 Kerry to +2 Bush in Ohio, and came within a hairsbreadth of flipping the result in New Hampshire.

The implications of that should be immediately obvious.  If this did indeed represent an actual error in the vote totals reported by the paperless electronic balloting systems in those two states, and the exit polls actually accurately represent the number of votes cast, then those two states should have been won by Kerry rather than by Bush.  If we refer back again to electoral-vote.com, we see that Florida has 27 electoral votes, and Ohio 20.  A little arithmetic shows that reversing the result in either of those two states is sufficient to give Kerry the election, and if both should in fact have gone to Kerry, this would change the election outcome from 279 Bush, 252 Kerry -- a victory for Bush by 27 electoral votes -- to 299 Kerry, 232 Bush, a victory for Kerry by 67 electoral votes.

Now, what does this tell us about the correct outcome of the election?  The truth is, nothing definite... it's suggestive at most, because there's no paper trail, no way to audit the result.  But one thing is unequivocally and undeniably clear from that very fact:

We absolutely must have that paper trail in future elections.  We cannot afford not to have that ability to audit.

This should not be negotiable.  The situation of having just held an election and being unable to audit or verify its result should never have been allowed to come to pass.  We can never know whether the electronically tallied vote totals are correct or not, and we can never recount them.  All we can do is take the manufacturers' word for it that there are no hidden bugs or backdoors in their code (on machines in which multiple, frequently trivial to exploit, vulnerabilities have been found), and that the vote totals as reported were correct.

This is simply, flatly, unacceptable.

Comments

Thanks for posting that

[jmkelly.livejournal.com]

To: Senators Feinstein & Boxer; Representative Barbara Lee

Did electronic voting threw the last election? It wouldn't have been hard; in a close election, just a slight electronic thumb on the scale in either of two states -- say Florida, or Ohio -- would do the job. And because many electronic voting machines leave no paper trail, and their programming is kept secret by their manufacturers, we have no way of knowing whether that happened.

But there is a strange set of discrepancies in exit polls: in counties that used paper ballots or electronic ballots with a paper audit trail, exit polls matched actual votes to within the margin of error; but in counties that used electronic voting with no printed audit trail, exit polls consistently differed from actual ballot totals by about 5% -- in Bush's favor. These descrepancies are illustrated dramatically at http://home.infionline.net/~arm3/exitPolling.pdf.

In Florida and Ohio, 5% was plenty. And Florida and Ohio gave Bush his victory. And it's one thing for exit polls to vary from actual votes, but for exit polls to vary from actual votes depending on the method of counting the votes is another thing entirely. As Thoreau said, "Some circumstantial evidence is very strong, as when you find a trout in the milk."

An investigation is called for, and tightening of election standards. Voting-machine programs should be made public, or at least reviewed by a reputable non-partisan panel of programmers. At the very least, paper trails should be required from all voting machines.

[arasirsul.livejournal.com]

Unfortunately, what I haven't seen, and what the people pushing this information at us seem reluctant to tell us is: just how much difference was there in exit polls in these states in previous elections? What kinds of differences in exit polls compared to results does one usually see?

While I would agree that being able to audit an election is a good and necessary thing, I worry when people start throwing stuff like this around, the "Oh my god, how could this possibly be? FUD!" . What are they really trying to get out of me?

-JDF

[ilcylic.livejournal.com]

Right. But it can't possibly hurt to have the extra data path.

-Ogre

[unixronin.livejournal.com]

Unfortunately, what I haven't seen, and what the people pushing this information at us seem reluctant to tell us is: just how much difference was there in exit polls in these states in previous elections? What kinds of differences in exit polls compared to results does one usually see?

Well, do the research. Election results from the 2000 election are available, as are 1996 and previously, and it's probably possible to find the exit poll results archived somewhere.

As has been pointed out elsewhere, though, polls are known to be an approximation at best; what's interesting is not when exit polls vary from actual results, but when exit polls vary from actual results in a consistent pattern that depends on the method of voting. And if there's no audit trail, there's no way to tell whether the numbers were cooked or not. Many races in this election that have turned out to be pivotal or potentially pivotal cannot be recounted because there is no record of the actual votes cast, only of the totals. And that means there's no way we will ever know whether there's actually substance to it and some of the machines were hacked, or whether it is indeed merely FUD.

[tquid.livejournal.com]

I may have mentioned this before, but what I personally would really like to see is a Constitutionally empowered Elections Commission. It works in India, which has a billion or so people, way, way, way more complicated voting with lots of minority parties. Oh, and they don't use automated machines, most places, or possibly any. I believe this would be a much more lasting and effective solution to all these various problems.

[unixronin.livejournal.com]

I could definitely go for that. An independent, neutral Elections Commission, not beholden to either party (but answerable to the Supreme Court) would definitely be a huge improvement over the parties themselves conducting the elections.

[fruitylips.livejournal.com]

There is also a slight problem with, for instance, calling Ohio a non-paper ballot state. Earlier this year, Ohio authorized 31 of its 88 counties to use electronic voting machines. Several apparently waited. However, in any case, far fewer than half the counties in Ohio used electronic voting machines. The rest used paper ballots like they always have. Which counties were the results differences found in Ohio? Did that correspond to the ones where the electronic voting machines were being used or the paper ballots?

Given that this is now the third election in a row where the exit polls failed to match the actual results, one has to wonder how much longer the networks and other folks who pay for them will continue to do so. Given the poor nature of the question about 'What was the most important issue for you in this election?', I've no trouble believing that the exit polls themselves are screwed up. You're right, tho, that there's way more required to really have any idea what is going on here and much of that info isn't and can't be available given how things were done. Also, how do the exit pollsters deal with early, mail and absentee voters?

[unixronin.livejournal.com]

My assumption would be, "They don't."

[fruitylips.livejournal.com]

Mine too, but I was hoping someone knew better. If that's really the case, tho, with the large numbers of early voters and absentee ballots cast now, it would seriously call into doubt the exit polling in the first place.