Please give me another explanation for this.

[solarbird]

Polling in the last few days before the primary as well as exit polling showed Senator Obama winning the New Hampshire primary over Senator Clinton. Exit polling also showed Senator Edwards in a solid third place.

Senator Edwards's finish statewide matched the polling. Senator Clinton and Obama's didn't. Various theories have been floated as to why.

Andrew Sullivan pointed to this analysis, which provides an extremely disturbing data point. In New Hampshire, some precincts are hand-counted. The others are Diebold machines. In the hand-counted precints, Senator Obama's and Senator Clinton's counts matched the exit and pre-primary voting, with Senator Obama winning strongly. In the Diebold-counted precincts. Senator Clinton won by significant numbers.

Please give me another explanation for this discrepancy. Please. Because this is a disaster. If it's Clinton's machine doing it, it's desperately stupid to have done so so sloppily - even if they trust the useless, fawning media not to talk about it. But frankly, I think it's too obvious and accordingly too stupid. On the other hand, if it's the GOP's - and if they've covered their tracks too well for it to be tracked back to them - it's a master stroke, setting up the Democratic party for internal war. In either case, if true, it destroys whatever confidence might have been remaining in the American elections system.

I have said many, many times that Diebold machines are trivially hackable fraudboxes that need to be banned. They need to be gone. This has to happen. Now.

ETA1: Petition to that effect here. It's not enough. But it's a start.

ETA2: At least one other report (see comments) says that the exit polling was within margin of error of final results. This does not match what I saw beforehand and the morning of the primary, but I didn't pay attention throughout. It also does not match what I read in coverage afterwards, but without the actual raw data in front of me...

ETA3: Here's the breakdown of the spread: GOP, which shows a strong machine-count gain for Mitt Romney, and Democratic, which shows a strong machine-count gain for Hillary Clinton at specifically the expense of Barak Obama, primarily in small and medium-sized towns, which are broken out. Also, see my comments here.

ETA4: Regarding to the media report mentioned below (and above, in ETA2); the report does not actually claim that the exit polling matched the machine counts. It notes that there were large numbers of undecideds and unsures going in to the election, and that Obama got his pre-election polling percentage, and that therefore the result is reasonable. It doesn't really address exit polling discrepancies at all. It further states a few key facts: 1) Exit polls said undecideds were split between Obama and Clinton. However, 2) as noted above, Obama got his pre-election polling numbers - almost exactly. (Also, Edwards also got his pre-polling numbers, iirc. This is not mentioned in the report, and I could be remembering wrong). This implies that all undecideds went to Clinton, and either lied about it to exit pollsters or undecideds switching to Obama exactly balanced out Obama's losses of support shown in earlier polls.

Comments

[kathrynt.livejournal.com]

According to Keith Olberman tonight, the results were all within the margin of error of the polling, and the Diebold machines are in precincts that are demographically different than the hand-counted precincts, and which demographics have been shown to favor Clinton.

[avram]

I've got one big complaint with that article:

The only difference between these two groups is that one had their votes counted by hand, and the other by Accuvote.

We don't know that. Or at least, I don't, and I don't think you or Bill Noxid do either. If part of a state uses one kind of voting machine while another part uses a different one, there's probably a reason for it. Like one part is wealthier, or has a different population density, or something. There are any number of factors that could correlate with both a shift in voting and a difference in voting machines.

I agree that we have to return to paper ballots, just so that we can trust that there isn't trivial electronic hanky-panky. But I'm not convinced that there was a problem here.

[solarbird]

This contradicts pretty much everything I had read or heard on the radio, but as I do not and have not had raw numbers in front of me - ETA'd to reflect your report.

[solarbird]

Or at least, I don't, and I don't think you or Bill Noxid do either.
Bill Noxid may and may not. I certainly don't. You'll note, I'm asking for other explanations. If there isn't another one, and a good one, we have a serious fucking problem.

And yes, if you're wondering, I know more than enough about these machines that I do consider any election using them automatically suspect, for good reason.

[king-chiron.livejournal.com]

The Bradley effect has been widely cited, the idea being that people aren't willing to admit that they wouldn't vote for a black candidate to pollsters. The Diebold/handcount discrepancy is interesting but we're mixing and matching data, are there other differences between the counties besides their voting method (age or wealth differences for example)?

Now, what would be great would be if someone could show that the Obama/Clinton pre-election polls were similar across Diebold/handcount counties, then you'd have, if not a smoking gun, at least a warm one. Hopefully someone has detailed enough data to look at this.

[king-chiron.livejournal.com]

Exactly, and I wonder what would drive a county to use machines vs. hand-counting. It would be nice to think it was due to extra-careful elections officials in those hand-count counties, but I'm willing to bet it's more a matter or money or the size or the county or something like that.

[kathrynt.livejournal.com]

If the numbers at http://ronrox.com/paulstats.php are to be believed -- and while I see every reason why they could be biased, I also have followed up a few of them and found them accurate -- then Romney had an *even bigger* swing than Clinton following the same split. Over seven percent as opposed to over five.

[technoshaman.livejournal.com]

fsckers. There's reasonable doubt either way, but the idea that there are still Diebold machines out there, and that the GOP has no qualms about doing this, is busted beyond all recognition. Because the one sure way to guarantee that we get an Elephant in the Orifice is that Hillary is the D's candidate. I honestly think they'll do something drastic to avoid that.... and I won't vote for her either.

Frankly, it might do well to let it leak to the Karl Roves of the world that if there is so much of a *whiff* of fraud, that they will end up with their heads on pikes....

[solarbird]

Swing in question is hand-counted vs. machine-counted, I add for the sake of clarify.)

It's a little difficult to compare here, because in the Republican case (according to this site), Romney was a huge beneficiary (+7.5%) with no single individual loser showing a drop of more than 2.7(ish) percent. With Obama and Clinton, it's a very clear shift, with no other Democrats losing noticeably - Clinton +5.4, Obama -3.0(etc), for a total of 8.4% swing. I think in the case of a swing this targeted, you have to consider both numbers. BUT there are plenty of reasons to call bullshit on that.

What's also interesting is that in both cases, the Party Establishment candidate benefited. Strongly.

The Democratic Party page on the same site breaks out hand-counted vs. machine-counted by town size, which gives us a look into the demographics question. In all cases where there were significant numbers (by percentage) of both hand-count and machine-count available, Clinton gained significantly by machine count. So large-town vs. small-town does not seem to be a factor, compared to machine-count vs. hand-count. (Small town swing to Clinton, machine count vs. hand count: 8.3%. Medium-sized town swing to Clinton, machine count vs. hand count: 9.9%.)

(In large towns, where only a tiny fraction of votes were by hand, Obama gained. But at that point you're dealing with hand-count numbers small enough small total changes make large percentage changes, which skews the data, IMO. This may of course be disputed if you like.)

[solarbird]

We've got one now: town size. Which didn't matter in sizes where there were comparable numbers of machine-counted and hand-counted votes; small towns and medium towns all showed substantial swings to Clinton in machine-counts vs. hand-counts. See ETA3 in original post, and more comments here (linked to in ETA3).

[kathrynt.livejournal.com]

My tinfoil hattery is that it's the GOP swindle on both sides. They swindled their side smart and the other side dumb, knowing that the dumb swindle would draw fire and let their own smart swindle escape inquiry.

[cafiorello.livejournal.com]

Here's Andrew Kohut's explanation in the NYTimes:

http://www.nytimes.com/2008/01/10/opinion/10kohut.html

Summary: Poor, rural white voters who are less likely to talk to pollsters are also less likely to vote for a black man.

Plausible, but then why would they be more likely to vote for a woman?

I did post a comment on the site about Diebold. We'll see what turns up.

[llachglin.livejournal.com]

Obama got a percentage of votes within the margin of error of several polls. What was unusual was that Clinton got more votes than predicted. On the other hand, if you include the large number of undecided voters then the obvious suggestion is that the undecideds broke heavily for her.

My understanding about the discrepancy between hand-counted and machine results is that while there was a couple of percent difference, Clinton still outperformed her polling in hand-counted precincts by several percent. Most of the surprise in the results was not the result of this small discrepancy in the machine count.

The theory about demographic differences was not Olberman's but mine, though it was just a guess. A correlation between higher Clinton votes and machine-counting is more likely to be due to a third factor than anything nefarious.

I wouldn't trust Sullivan to report on this, by the way, as he has a longstanding grudge against the Clintons. Who knows why, but he's not a neutral reporter. He's a gay conservative. Clinton is probably my third favorite Democratic candidate, but I'm getting sick of the anti-Clinton smearing going on out there, from all sides. (And while I don't want her to win the nomination, I'm happy she won New Hampshire after some of the sexist comments by the Edwards and Obama camps and much of the media. It was important for that to be repudiated.)

[llachglin.livejournal.com]

Oh, and the Wilder/Bradley/Dinkins effect is NOT what was happening, because Obama got pretty much the numbers that polls predicted. What was off were
the predicted numbers for Clinton. It seems like soft Edwards leaners switched to Clinton, along with many undecideds. Obama didn't get the undecideds but DID get the percentage he polled. Therefore, no covert racist Wilder effect.

This also argues against the idea that Clinton's camp or someone else was someone manipulating Diebold machines by adding additional votes for Clinton, as that would have lowered Obama's percentage below his polling. The numbers suggest a swing from Edwards and undecided, and nothing else.

[llachglin.livejournal.com]

From Daily Kos, a further explanation why this whole theory is garbage:

http://www.dailykos.com/storyonly/2008/1/10/02623/2264/85/434176

Summary: ALL votes in NH are by paper ballot, so any manipulation in the counting would be easily discoverable. It's similar to the system in King County, where people vote on paper but computers are used to tabulate. There are zero touch screen voting machines in NH, so no chance of undiscoverable hacking.

[solarbird]

I am pleased to note that Representative Kucinich has decided to ask for a recount on the basis that this statistical variance - a variance significantly larger than Florida in 2000 and Ohio in 2004 - is in fact worth checking, since, as you note, there is a paper trail. I feel that this is the appropriate response, given the history of outcome-altering election tampering in recent presidential elections, and will hope that the New Hampshire hand-count does in fact validate the machine count.

[llachglin.livejournal.com]

I agree. Any hint of something weird should trigger a recount, and I commend Kucinich for calling for one. I actually think it's worth the expense just to hand-count every time and make the confirmed hand-count the official results every time in every election anywhere.

[solarbird]

Also, a similar suggestion from Florida 2000 turned out, upon study, not to be true at all.

What I'm looking for is a researched demographic explanation that addresses the actual data; alternatively, I'll take the recount Rep. Kucinich has requested. But even assuming nothing is going on, letting an 8.4-point swing like this (combined) just rest unchecked simply is not acceptable.

[mojave-wolf.livejournal.com]

You beat me to the punch on mentioning the (several) NH dems on Kos who said the voting patterns for the various locales were historically pretty much what you'd expect; plus, as some others noted, w/the paper trail there, and the real game yet to come, this one would be *stupid* to hack (unless you think Clinton's corporate-friendliness is greater than Obama's, which seems to be the public perception but isn't borne out at all by a study of their records and friendships, where both are a bit disturbing, tho she's a lot more open & honest about it)(can you tell I like Hillary better? whereas Obama keeps losing me more by the week, both from studying his record and his campaign's sexism, which is just a more subtle form of what the MSM is doing.)

& seriously, practically everyone I've heard comment on this, including a few who support Obama, thought Hillary kicked ass in the debate, he didn't do so well, and then there were all the people pissed off on her behalf about the non-crying incident & the ever more obvious damned if you do, damned if you don't coverage of her, & I totally see nearly all the last minute undecideds breaking for her, just as all the people suddenly breaking for Obama because of Iowa made sense . . .

All that said, I'm with the two of you (and opposed to most of the Koos people) in agreeing that, hell, you have something that looks weird at least on first glance, you have a paper trail, check it out . . .