A Plea.

[pmb]

Despite my disappointment at the outcome of this election, what is most important to me is that the people have at least nominally spoken. The election has been held, and we support the system, even if we do not support the winner. At all.

To that end, if faith in the system is lost, then all is lost. Because you can't fix anything within a broken system. So please, please, PLEASE, someone tell me where the fault in this graph lies. Because I desperately want to believe that he actually won. The alternative is too awful.

Comments

[jes5199.livejournal.com]

either way, here's a problem with the system: without an audit trail, it is impossible to prove that those graphs are a hoax.

[pmb.livejournal.com]

Absolutely true as far as an audit trail goes.

But those graphs being a hoax? Not at all - as long as I can get my greasy hands on raw exit poll data and on the reported totals in each state and county, then I could reproduce them. What I really need to make myself "happy" about this election would be raw exit poll data (or at least "un-touched-up", because they started combining exit polls and actual vote data at the end there) as well as the reported totals in each county. Then, if I could not reproduce that graph, I would put it down to some random lefty angst causing someone to lie and use bad data. Otherwise it remains convincing evidence of a stolen election. Which I do not want at all.

[tdw.livejournal.com]

Is there any corroborating evidence of that? I don't trust random blogosphere memes.

[pmb.livejournal.com]

Well, the opposition seems to also be calling for an investigation of why the exit polls were so wrong, based on the idea that the election was right and th epolls were wrong - http://www.thehill.com/morris/110404.aspx

So at the very least everyone can agree that something is fucked.

[katiemouse.livejournal.com]

If exit polls are what people said they voted, they might have lied. I have a friend who voted for Bush but if anyone he doesn't know asks(in Democratic Seattle), he says he voted for Kerry but he doesn't want to deal with angry people who disagree with him.

[leech.livejournal.com]

I'd look at the timing and methologies of the various exit polls first. Democrats vote earlier in the day than Republicans.

[pmb.livejournal.com]

The only data I could find was on the Florida one - late exit polls (i.e. taken late in the day) suggest the numbers shown on the graph above.

[greyowl.livejournal.com]

Also, I doubt absentee or early voters are included in the exit polls. My county (Bernalillo, NM) is probably unusual, but over half of us voted early or absentee. And if you look at the unofficial results (http://www.bernco.gov/upload/images/clerk/general04/resultsPage1.html), which the county breaks down into precinct/absentee/early, you can see some definite differences. (In this particular case, absentee voters were more likely to vote for Kerry.)

[mbrubeck.livejournal.com]

1. This is not statistically significant. You are looking at a population of 10 states with electronic voting, and only 2 (!) with paper ballots.
2. We should expect exit polls errors to vary systematically from state to state. Exit polls use non-uniform sampling . Not all counties, polling places, and times of day are represented equally. The accuracy of the poll depends on the accuracy of the pollster's demographic and turnout models, which in turn will vary state by state.
3. The vertical scales have been manipulated. If you adjust the vertical range for Wisconsin (paper ballots), you can get a graph essentially identical to Pennsylvania (electronic voting) and New Hampshire (electronic voting).

[agthorr.livejournal.com]

I second this. There's no way to know that the person constructing that image didn't and pick and choose what states to include. We need to see all the states, and the y-axis should be the same on all the graphs.

[leech.livejournal.com]

Weren't those all swing states?

[pmb.livejournal.com]

1) Agreed. More data is essential. This is not conclusive, but is suspicious.

2) Exit polls are traditionally used as the primary means of investigating whether or not a count will be honest. We've had 50-odd years to hone our methodologies. And everybody got it wrong in the exact same way this time? Too fishy.

3) Ummm. I'm not seeing that. I agree that the vertical scales have been manipulated, but the primary point of interest for me is the discrepancy between the exit polls and the tabulated votes. Wisconsin (paper) went from a exit polled 52/48 win for kerry to a 51/49 win for kerry. That intuitively seems well within the margin of error. Pennsylvania (electronic voting) went from an exit polled 60/40 win for Kerry to a tabulated 51/49 win for Kerry. 9% is way outside any intuitive margin of error. New Hampshire went from a 41/57 win for Kerry to a 50/49 win for Kerry. Again, 7% is way outside any reasonable margin of error.

Doing some quick math... the exit polls were off by an average of 1.5% in the paper ballot states shown, and by 5% in the electronic voting states shown. That's 5% for each candidate - which means they got the spread wrong by an average of 10%, as opposed to 3% in the paper states. Something was systemically wrong with the polling or the voting. And we've been polling in basically the same way for half a century, while the electronic voting machines are pretty new.

[lucasmembrane.livejournal.com]

I'd like to find the raw data and do some more in-depth statistical analysis. In addition to the factors you mentioned, there's also the fact that the polls were *consistently* off in favor of Kerry. If there's a problem with the sampling technique, some polls would favor Bush and some would favor Kerry.

Er, I guess you mentioned that in point 2. But yeah. Fishy.

Damn, I need more free time in my life for projects like this. And I need to find my old statistics book.

Anyway, if you find a good source of buttloads of raw data, point me to it. I may be able to throw something together.

(And as far as giving you hope is concerned, I'm afraid I can't help. I fully believe that the election was stolen, and, lacking an audit trail, that we will never know for sure.)

[olstad.livejournal.com]

I think you're moving a bit too fast on the 50-year validation of exit polling. You can't just take an SRS of voters; you don't have access to the whole population at once...
a) We did have record turnout this year (and from different ages/groups)
b) Most Americans feel _very_ different about exit polling than they did during the last pres. exit poll
c) Even during the campaign, we noticed that different organizations (notably, Gallup vs. whoever MoveOn thought was better) got notably different answers on polls, apparently due to different survey architecture.
d) Some of the voting rules changed this time (e.g. provisional ballots in all states-- it is possible this changed who went to which polling places).

All that said, I'd be much more comfortable if we could get some checkable data here...

source validity

[cosyne.livejournal.com]

for what it's worth, that does seem to be posted on the same TLD that brought you goatse... on the other hand, you can find some scary stories about voting machine malfunctions if you look.

[mbrubeck.livejournal.com]

By the way, here's a real statistical analysis of Florida that compares counties with paper ballots to those without. As the evoting-experts group points out in a later post, we need more real analysis by trained statisticians, to figure which effects are significant and what they could mean.

[roninspoon.livejournal.com]

As I understand it, exit polling is extremely flawed and innacurrate. It's really only good for the media who need feedback to prepare headlines and pump us full of hope and or despair depending onthe situation. Having said that, I think there is increasing evidence that the system is being gamed by each side to take minute advantages of the errors and loopholes of the election process.

As an aside, if you were going to rig an election, it's a more plausible lie (and more difficult to challenge) if you excell inside the margin of error, rather than by a landslide.

[cabanasloth.livejournal.com]

I'd agree that something has gone extremely screwy with the exit polls. This article [slate] claims that exit polls have historically always been wildly inaccurate, and that a ten percent margin of error is well within reasonable expectations for poll validity. However, I have yet to see anything anywhere that addresses what appears to have been the systematic pro-Kerry bias in the exit polls. It may be that Democrats tend to vote earlier in the day, as suggested above, but that seems a bit strange to me - why, exactly, would Democrats vote early? The only plausible idea I've been able to come up with is that exit polls may oversample urban precincts, which would certainly introduce a systematic bias like the one we seem to have here. However, that would seem like a *really* obvious flaw in the polling design, and I have trouble believing that the polling organizations would not have taken this into account.

This whole story seems very strange to me. The media story on this seems to be just that "exit polls suck", and they're leaving it at that. But I've yet to see a reasonable explanation for *why* they are apparently so bad - not just imprecise, but biased. I don't really expect to see such a story, either, since the difference between sampling error and bias may be a bit too tricky a concept to convey in the press. But it's frustrating and not satisfying at all.

[kirin]

I've seen that set of graphs before. I'm somewhat suspicious of the generalizations going into it, given that I voted in NC by paper ballot (optical scan). I have no idea what parts of NC might have been electronic.

(We also had tons of early voting around here... I voted in the middle of October.)

[dragonmudd.livejournal.com]

Exit polls are not random. I was listening to some "experts" talk about why the exit polls were so different, and they gave some acceptable reasons. I can't remember most of them, but one was in Florida. Most of the exit pollers were english speakers. Many of the voters spoke english poorly or not at all. In such a case, the exit pollers would talk to fewer of the spanish speaking voters than the english speaking voter. Additionally, they say that the hispanic community is more likely to vote republican, so that would be an example of how the exit polls can skew the data.