Berkeley Researchers Analyze Florida Voting Patterns

empraptor writes "Researchers at UC Berkeley have crunched numbers and determined that 130,000-260,000 excess votes went to Bush in Florida. They have held a conference and posted their findings online. You can find articles on their research from CNet, Wired News, and many other sources. While the research used statistical analysis based on past elections and demographics, how else do you verify that a paperless voting system is working properly?"

Some thoughts

[daveschroeder]

http://abcnews.go.com/WNT/story?id=239735 [go.com] (video [wisc.edu])

Doug Chapin, a nonpartisan election analyst, finds the claims to be baseless. "There were no problems that would lead me to believe that there were stolen elections or widespread fraud," he said.

"There was no overwhelming reason to cast doubt on the outcome of this election," seconded Democratic strategist Donna Brazile, the campaign manager for Al Gore's 2000 campaign. "George Bush got more votes this time."

http://www.boston.com/news/nation/articles/2004/11 /10/internet_buzz_on_vote_fraud_is_dismissed/ [boston.com]

Much of the traffic is little more than Internet-fueled conspiracy theories, and none of the vote-counting problems and anomalies that have emerged are sufficiently widespread to have affected the election's ultimate result.

Kerry campaign officials and a range of election-law specialists agree that while machines made errors and long lines in Democratic precincts kept many voters away, there's no realistic chance that Kerry actually beat Bush.

''No one would be more interested than me in finding out that we really won, but that ain't the case," said Jack Corrigan, a veteran Kerry adviser who led the Democrats' team of 3,600 attorneys who fanned out across the country on Election Day to address voting irregularities.

''I get why people are frustrated, but they did not steal this election," Corrigan said. ''There were a few problems here and there in the election. But unlike 2000, there is no doubt that they actually got more votes than we did, and they got them in the states that mattered."

''I think it's safe to say that on the votes that were cast in Ohio, Bush won," said Dan Tokaji, a law professor at Ohio State University who is working with the ACLU to challenge Ohio's use of punch-card ballots. ''If the margin had been 36,000 rather than 136,000, we would have seen another post-election meltdown."

http://www.sacbee.com/state_wire/story/11436220p-1 2350492c.html [sacbee.com]

All three said their networks had set up investigative units to review any claims of voter fraud or problems with electronic voting technology this year, but that nothing significant had appeared anywhere to affect the election's outcome.

"A lot of the allegations we've looked into, they're just not true," Shapiro said. "Believe me, I'd love a juicy story about the election as much as anybody. Florida was a great story, but it's just not there this time."

A frequent charge levied after the 2000 election was voter disenfranchisement and ballot spoilage due, in large part, to antiquated, malfunctioning, or broken mechanical voting equipment. Legislation was introduced guaranteeing a minimum standard for the equipment and processes associated with voting in all jurisdictions. Since we are living in the 21st century, electronic systems were specified. $3.9 billion was set aside under HAVA to replace all mechanical punch card systems with electronic systems by 1 January, 2006. The goal is to ensure a consistency and fairness in the appearance and operation of the voting systems, both for voters and local election officials.

After the 2000 presidential election, Congress passed the Help America Vote Act of 2002 (HAVA) [fec.gov]:

To establish a program to provide funds to States to replace punch card voting systems, to establish the Election Assistance Commission to assist in the administration of Federal elections and to otherwise provide assistance with the administration of certain Federal election laws and programs, to establish minimum election administration standards for

Re:Statistical?

[arodland]

Read the abstract, if not the actual paper; it's a little deeper than that. It says that Bush got more votes than expected, and that the counties where he got larger-than-expected numbers of votes are the same counties that used electronic voting, to a statistically significant level.

Just fix it! Support the bills that will!

[daveschroeder]

A frequent charge levied after the 2000 election was voter disenfranchisement and ballot spoilage due, in large part, to antiquated, malfunctioning, or broken mechanical voting equipment. Legislation was introduced guaranteeing a minimum standard for the equipment and processes associated with voting in all jurisdictions. Since we are living in the 21st century, electronic systems were specified. $3.9 billion was set aside under HAVA to replace all mechanical punch card systems with electronic systems by 1 January, 2006. The goal is to ensure a consistency and fairness in the appearance and operation of the voting systems, both for voters and local election officials.

After the 2000 presidential election, Congress passed the Help America Vote Act of 2002 (HAVA) [fec.gov]:

To establish a program to provide funds to States to replace punch card voting systems, to establish the Election Assistance Commission to assist in the administration of Federal elections and to otherwise provide assistance with the administration of certain Federal election laws and programs, to establish minimum election administration standards for States and units of local government with responsibility for the administration of Federal elections...

The putative reasoning for going with electronic systems was likely that since we have managed to design accountable and reliable electronic and computing equipment for the management of our power, medical care, money, etc., it likely was more or less assumed by the legislature that such accountable systems could also be applied to voting.

A bill has been introduced to amend HAVA. H.R.2239 [loc.gov] and its twin Senate counterpart S.1980 [loc.gov], discussed further here [verifiedvoting.org], will amend the Help America Vote Act such that there is "a voter-verified permanent record or hardcopy" attached with each and every ballot cast by every voter, and that "any voting system containing or using software shall disclose the source code of that software to the Commission, and the Commission shall make that source code available for inspection upon request to any citizen".

Additionally, the three electronic voting manufacturers already have the ability to add permanent, individual voter-verified paper audit trails to their products. Some e-voting critics make it seem like vendors are resisting. However, it is the local election boards that are resisting (as well as the slow march of bureaucracy). The e-voting vendors will build - and sell - whatever municipalities will buy.

Re:Ohio would be better

[Megaweapon]

It has been a stuanch Democrat state

Huh? I live in Ohio and I can tell you it isn't "staunch" for the Democrats.

Re:Paper trail not enough

[beeplet]

The point of paper receipts is that you don't keep it, you verify it and then put it in a regular ballot box. That way if there's any question about the electronic tally, there can be a manual recount.

Mirror

[Rufus211]

site's getting slow (8k/sec), so I've mirrored it:
http://www.contrib.andrew.cmu.edu/~pnelson/ucdata. berkeley.edu/new_web/VOTE2004/ [cmu.edu]

Re:Paper trail not enough

[morcheeba]

The point of paper trails is that you keep them -- if there are any questions later, you count those by hand and see if the machine is correct. And, like other have said, you can't tell if the code running on machines is actually the publicly-available source you can view (even if you compile it yourself on the machine before you vote - the compiler could be tainted)

Re:Why Berkeley?

[eyeye]

All schools are liberal because intelligent people work and visit there.

Dumbasses are generally republicans.

Kerry Could still Win...sort of.

[JavaLord]

Not sayin' it is so... but HAD the election been accidentally given to Bush, now that Kerry has conceded, what would the legal recourse be??

Bush hasn't really won until the electoral college vote is done, which I believe is in December. If Kerry won a court battle in Florida the electoral votes could still go the other way. It is up to the free will of the electoral represenatives. The point of this article is moot anyway, even if Kerry got 260,000 extra votes, it wouldn't matter, Bush won Florida by about 400,000.

Re:A legal question

[Rufus211]

conceeding means nothing legally, it just means "ok, I'll shut up now." The only thing that matters legally is what the electoral college votes. According to wikipedia [wikipedia.org]:

Each state's electors meet in their state capitals in December , 41 days following the election, at which time they cast their electoral votes. Thus the electoral college never meets as one body. The electoral votes are then sealed and sent to the President of the Senate (i.e. the sitting Vice President of the United States), who retains them until the new Congress convenes in January. At that time, the votes are opened and counted in the presence of both houses of Congress. The candidate who receives a majority of electoral votes for President becomes President, and the candidate who receives a majority of electoral votes for Vice President becomes Vice-President.

So we have about a month for the electoral college to change its mind.

Not published.

[helix400]

I would like to point out one extremely important fact. This paper was not published in a journal. This was a simple statistical analysis by four students.

Let the scientific method work this out. If a paper has merit, let it be analyzed by stastic professionals, and if it does have merit, any statistical journal would be happy to run a major news story that would give them publicitiy.

But too many of these wannabe statiticians are not publishing their results. They make unrelaistic assumptions, they use questionable approaches to making claims, they don't use enough variables (in the case of this report, they didn't even factor in Nader!) And when they find something they believe is significant, they bypass the scientific method completely, and rush straight to internet blogs or PRNewswire.

Again, let the scientific method take its course, and be very cautious of anything that doesn't.

Re:They're never going to give up, are they?

[PktLoss]

RTFA They are NOT suggesting that Kerry won, they admit rather early in each of the texts I read that the vote descrepancy was not enough to change the result of the election.

They are over it.

Re:A legal question

[Matt Amato]

This is where people fail to realize how the electoral system works. First off, no one has been elected president yet. States have simply held a poll to let thier electors know who people want them to vote for. The Actually election doesn't happen until the Monday following the second Wednedsay in December. If enough electors decided to switch sides and vote for Kerry, then he would become President. Kerry conceding doesn't mean he can't get elected, it just means he admits that he has lost. If it turns out that that isn't true, or if the electoral college changes it's mind, then he becomes president. Just like if the electors decided all to vote for Howard Stern, even though he isn't running, well, then he's president. See http://people.howstuffworks.com/question472.htm for more info on the electoral system.

Re:Why Berkeley?

[pherris]

Because Berkeley is one the best schools in the world. Yes, there are "Berkeley" politics in and around the community but for the vast, vast majority of the time their scientific research is nonpartisan and beyond reproach.

Re:A legal question

[jdavidb]

Concession has absolutely no legal implications. It just means you'll quit putting any resources into contesting the election. (i.e., you won't ask for recounts, you'll tell your supporters it's over and you fought a good fight, etc.). If you remember, Al Gore actually conceded and then took his concession back. :)

The election still is not final. Each state will certify its election results through its own legal process which may differ from state to state. I don't think that's complete in most (any?) states, yet. The popular vote counts are only final when that certification process happens, so if recounts started showing up for Kerry, the election could still change.

Even after that, the electoral college doesn't meet for awhile, and it's not over until they vote.

MOD PARENT UP

[SlipJig]

It's about time somebody jumped on all these "liberal bias" claims. The "bias" argument is a distraction, and an excuse to avoid actually having to prove what you're saying.

See Carl Sagan's Baloney Detection Kit [gatech.edu] for a good description of logical fallacies, including the one mentioned in the parent post.

Re:Utterly Believable

[Yeechang Lee]

Bush and the GOP appear to have used Diebold voting machines to steal yet another election

Neat trick, considering that no Florida county uses Diebold e-voting machines [wired.com].

Re:Paper trail not enough

[dustman]

This can still be messed with. When the binary is read (to do the byte-compare, or to compute a digital signature), the machine can return the "correct" binary, while it was actually running a hacked binary.

You still need a paper trail. The ballots can be counted, doublechecked, you can use 2, or 3, or N groups of people with whatever affiliation. This is where vote security lies.

I live in New Hampshire. When I voted, my local voting place was using a Diebold optical scanner vote box. Just like any other paranoid slashbot, I don't trust Diebold. But I do trust optical scan ballots, because when the numbers become contested, real people can do a recount.

Re:A statistical analysis proves exactly what?

[arodland]

You should try understanding the premise before you go driving things through it.

What it says is "Here's the differences between actual tallied votes and what various other models predict. Notice how the trends work one way in certain counties, and a very different way in certain other counties. Notice how these counties coincide very well with the counties that used e-voting. In fact, there's less than one chance in a thousand that it could have happened that way randomly."

So it shows that there's a marked increase in Bush support in e-voting counties. That doesn't prove that fraud is the cause, but if you wanted to disprove it, then you would be wise to come up with some other reason why those sets of counties happen to overlap so well. And you would be wise to note the factors that they already corrected for.

Re:Possible explanation -- the values voters

[learn fast]

No.

RTFA [berkeley.edu]:

Compared to counties with paper ballots, counties with electronic voting machines were significantly more likely to show increases in support for President Bush between 2000 and 2004. This effect cannot be explained by differences between counties in income, number of voters, change in voter turnout, or size of Hispanic/Latino population.

Re:The REAL red flags in this debate

[Ironsides]

I actually asked the head official for my area why they don't go to a paper backup system. The completely rational answere is that they believe that no one would ever trust the initial computer count. Basically, here is the rundown of events they believe would happen if there was a paper trail.

1) Everyone votes
2) Electoronic machines report numbers
3) Losing party sues to have paper trail recount
4) Paper trail recount confirms computer count
5) Repeat 3 and 4 a few times.
6) Repeat in every election these machines are used in.

Given that the votes are challeneged and recounted every time, there is no point to going electronic at all. Also for those who care, my county is ~70% Democrats and votes that way. Although I do know that Illinois has many problems with Dead voters voting more than once. And they aren't electronic and they are heavily Democrat.

This Land Is Red Land, Paid For by Blue Land ...

[rfischer]

As students of the federal budget know, the citizens of some states pay more in taxes than they get back from Uncle Sam in grants and benefits. The late Senator Daniel Patrick Moynihan commissioned frequent studies that showed how New York was getting the shaft. Arnold Schwarzenegger was stunned to learn upon taking office that for every dollar Californians send to Washington, they get back only 77 cents--an imbalance that topped $50 billion in 2003.
linkus jucius [fortune.com]

Re:Not published.

[learn fast]

Yes, "too many of these wannabe statiticians are not publishing their results." I'm not sure who you're talking about, because if you RTFA (!!!) you will find that they put all of their data online. If you can find a flaw in that, then you've got something.

Journal-published articles take 6-18 months to be published. This article is only a working paper, and it should be treated as such, but it would be definitionally impossible for a academic-refereed article to be coming out within a relevant timeframe.

Re:And How About Mechanical Voting Machines?

[Anonymous Coward]

Actually, New Hampshire went to Kerry and there is a recount in the works focused on voting problems.

Re:Not published.

[schussat]

This paper was not published in a journal. This was a simple statistical analysis by four students.

As one commenter has already pointed out, the process of review and publishing takes at least a year, but it's common practice --- in all fields --- to circulate working papers and drafts. They've made all of their data and methodology available, so that appropriate review and replication can be conducted.

Also, the paper's lead author is a full professor, not a student.

-schussat

Re:Paper trail not enough

[Phisbut]

f everything is done by pen and paper, 1) it'll take forever to get the election results, and 2) the losing candidate can still question the vote.

1) Pen and paper doesn't take forever to count. In Canada, we can get federal election results approximately 3 hours after polls close, and they're all pencil and paper. Just because the US has a larger population doesn't mean it'll take longer, just hire more people to count and that's it.

2) If the losing candidate questions the vote, he and his opponent can both witness the recount. One can hardly question again after having witnessed the recount.

Correlation vs causality and 'loaded language'

[Sai Babu]

Irregularities does not have the same connotation as differences. Yet their "irregularities" is merely the difference in fit between what their model predicts and what they measured. Sure, you might call this an irregularity if you are fitting multiple data sets to a model and one of the sets doesn't fit. In this case, they have one model and one data set.

"Compared to counties with paper ballots, counties with electronic voting machines were significantly more likely to show increases for President Bush between 2000 and 2004. This effect cannot be explained by differences between counties in income, number of voters, change in voter turnout, or size of the Hispanic/Latino population."

* The effect cannot be explained within the framework of their model.

They state that their analysis takes account of:
- number of voters
- median income
- Hispanic population
- change in voter turnout between 2000 and 2004
- support for President Bush in 2000 election
- support for Dole in 1996 election

The counties with evote are the three largest counties. One should be careful in weighting the significance of the variables and data points. These three counties significantly skew the chosen fit.The change in turnout, past Bush support, and Dole support are not really relevent when comparing different demographics (three most populus counties with others).
The Hispanic vote is basically 'in the noise' outside of these three counties and there was a major shift in Hispanic vote between the Dole-Bush contest and the Kerry-Bush contest.
A little research will show that the shift in Hispanic vote was very significant in the 2004 election. Here is one of many sources [imdiversity.com]. Google will find many more.

Re:Paper trail not enough

[Lucidwray]

I have to totally agree with the 'pen and paper' comment. In my district here in Texas we have a great voting method. The ballot is a full size sheet of paper with each race and its canidates listed on it. Next to each one is a "Scantron" like area that you fill in with a black permenant marker. Once you have filled out the ballot you go stick it in the voting machine and it does the optical scanning. Voila! Super easy.

You get the advantage of a quick count from the optical scanning, plus you get the advantage that the actual paper ballot the voter filled out is stored in a secure steel box. If there is any question about the count you simply go back over the ballots.

No need for printing a paper trail, let the people create the paper trail.

Re:Two things

[Anonymous Coward]

Don't just RTFA, but look at their DATA!

According to their data, on average, the counties that used paper ballots increased their votes for Bush by .02%. The counties that used the "e-touch" as they call it increased their votes for Bush by .01%

Where's the conspiracy?

Ohio numbers don't match

[opencity]

29 precincts in Cuyahoga County, Ohio, reported votes cast IN EXCESS of the number of registered voters - at least 93,136 extra votes total.

official Cuyahoga County Board of Elections website [cuyahogacounty.us]

Re:Two things

[tbannist]

Blackboxvoting.org [blackboxvoting.org] says otherwise, and their most recent posting shows that there seems to be deliberate voter fraud going on. You don't refuse to show citizens the signed copies of the vote tallies and then try to give them ones that are unsigned that have different totals, and dump the original copies into the trash all by accident.

There's also this report and the report [ustogether.org] that shows a significant and consistent difference in voting patterns in counties using Diebold electronic scanning machines. That's three different sources confirming that something is wrong based on three different investigative measures. How different ways does someone have to show that the totals don't add up?

Re:Paper trail not enough

[Grishnakh]

Not just a little, but a lot. Here in Arizona, my ballot was about double the size of a letter-size sheet of paper, and was full on both sides of selections to make. These included President, Senator, Representative (all at the federal level), then state representatives, local representatives, local school board officials, then several huge columns of which judges we wanted to get rid of, and lastly all the statewide Propositions. All the results for all these different selections are tallied in one night. Why hire more people (who make mistakes easily) when machines can do it much faster and more efficiently?

The only problem with machines is verifying they're accurate.

BTW, do you do your laundry with machines? If so, why? Why not just hire some servants to do all your household chores for you? If you don't do this, or don't do all your household chores by hand with no machines at all, then you're a hypocrite.

Re:Two things

[Anonymous Coward]

The exit polls now posted on CNN were 'recalibrated' after the actual results were released. These new exit polls are completely meaningless. The original uncalibrated exit polls did NOT show Bush winning.

Same in Australia

[hayden]

All our elections are carried out with pencil and paper. Everybody votes in the same way and all the counting is done in a central location. We have problems with distance (if our largest electorate was a country it would be the 13th largest) so it takes a while before counting is done but everybody gets to watch and it's all verifiable.

there is an official and accepted reason...

[seibed]

Move along folks, nothing to see here.

http://www.cleveland.com/news/plaindealer/index.ss f?/base/cuyahoga/1100082911218052.xml [cleveland.com]

Re:Paper trail not enough

[CaptainCarrot]

That's easy enough if the ballot is a simple one. You couldn't possibly do it in many parts of the US. I worked as an officer at an election precinct this year, in a county where we use optically scanned ballots. After the polls close, part of our job is to sort through the ballots looking for those that are mismarked, have write-in candidates, or would otherwise be unreadable by a machine, and set them aside for hand counting by the Elections Board. With four people working a single ballot box, that took the better part of an hour by itself.

Now, besides the presidential election there were whole slates of local and state candidates, a US Senator and Representative to choose, and close on a dozen referenda. There is simply no way to do an actual count that quickly with a ballot that large.

In principle, I agree with you. But for a state to change its system to manual counting, it would require it to completely change the way it conducts elections. It's not a change we can simply drop into place.

Reasons for exit poll discrepancies

[Anonymous Coward]

In this case the exit polls showed that people were voting for Kerry but the counts showed otherwise.

Now what? How do we know which is true?

The president of the exit polling company, Warren Mitofsky, explained on the News Hour with Jim Lehrer on November 5, 2004, how the exit polling works and why it was imperfect.

The exit polls, in the words of Mitofsky, "interviewed almost 150,000 people nationwide on Election Day. We interviewed in every state but Oregon, since they don't have any people at the polling places, and we also interviewed a national sample of polling places."

You gonna allow a sampling of about 10% to determine the outcome of the entire election?

The full interview is available online (http://www.pbs.org/newshour/bb/politics/july-dec0 4/exitpolls_11-05.html [pbs.org]) in transcript form, streaming video, and streaming audio. In particular (bold emphasis mine):

TERENCE SMITH: Why did the early numbers show Senator Kerry ahead?

WARREN MITOFSKY: Well, Kerry was ahead in a number of the -- in a number of the states by margins that looked unreasonable to us. And we suspect that the reason, the main reason, was that the Kerry voters were more anxious to participate in our exit polls than the Bush voters. That wasn't the case in every state. We had a few states that overstated the Republican margin. But for the most part, it was Democratic overstatement for the reason I just gave you.

TERENCE SMITH: So you're saying that some Bush voters would come out of the polling places and simply decline to participate; if so, why?

WARREN MITOFSKY: Well, in an exit poll, everybody doesn't agree to be interviewed. It's voluntary, and the people refuse usually at about the same rate, regardless of who they support. When you have a very energized electorate, which contributed to the big turnout, sometimes the supporters of one candidate refuse at a greater rate than the supporters of the other candidate.

TERENCE SMITH: Well, if you thought those numbers were suspiciously high for Senator Kerry, couldn't you correct the sample, as you say in your business?

WARREN MITOFSKY: Well, we recognized the overstatement in the exit polls in mid-afternoon, and we told the members of NEP about the suspicions we had, which they chose to ignore. The correction, in this case, is to wait for the vote returns in those same sample precincts and use that for projections. There were no mistakes in the projections. We were very cautious with them, and none were wrong, even though the exit polls did overstate Kerry in a number of states.

TERENCE SMITH: Right. Now, this is supposed to be not-for-broadcast information as it's passed along to the organizations, but in fact, it affects their coverage and influences their thinking as they work on the coverage, and obviously someone leaked it.

Who? Who leaked it? How did it get out so widely on the Internet that, in effect, by mid-afternoon, it was public information?

WARREN MITOFSKY: The information is available to all the NEP members. That's five television networks and the Associated Press. It's also available to all the subscribers, which includes major newspapers and local television stations. Any number of people had access.

The reason we have this information at midday is so we can go over it, find any problems with the way we're displaying it, laying it out, any problems that might confuse us when the polls close.

I urge everybody to read the interview. Please mod this post up, so that people won't keep asking the same questions over and over.

Re:Vote Fraud Smoking Gun

[Izaak]

Sorry, Chachi, the preloaded votes were for Kerry. Widely reported in the news. Your other 'facts' are wishful germs of urban legends, as well.

Odd that all these 'wishfull germs' were reported by mainstream media and have not been dispuited by aanyone. For example:

In nine counties, electronic vote machines count Democrat votes as Libertarian. [indystar.com]

This relates to my earlier mention of a vote machine glitch tipping a local election. [pal-item.com]

As for those reports of extra votes [westchesterweekly.com], if this site can be believed, it might only be the tip of the iceburg [rise4news.net].

As for those pre-loaded votes being for Kerry instead of Bush, I could not find a news report that backs up your assertion, so I would welcome a reference if you have one. I will gladly admit I'm wrong on that detail if you do. Voter fraud should be a bi-partisan concern... we need to shine a spotlight on it regardless of which side is doing it.

And my original point is still valid. The reports coming out of Volusia county are the most damning evidence of voter fraud to come to light so far. This event was caught on video tape and witnessed by the police... so it is difficult to discount. It is certainly too early to shout that the election was definitely stolen, but this certainly raises serious concerns and undermines the confidence in the election for many. If the election was not stolen, then there is nothing to fear from an audit; it will simply restore confidence in the election process. How can that be bad thing?

Cheers,

Re:Vote Fraud Smoking Gun

[Izaak]

Does that discount the possibility of fraud in this election, or suggest fraud in the last 6 as well?

Actually, the statistical study done at Berkely takes that whole 'Dixiecrat' phenomenon into account. It uses past voting trends and compares them to the recent election. The big news is not that the votes don't match party registration, but that the big variations from past behavior occur only in the precincts using electronic voting machines.

Read this...

[nha]

http://politics.slashdot.org/politics/04/11/19/175 4249.shtml?tid=226&tid=126&tid=103

before deciding that machines are evil and paper trails are the answer. The article shows how the vote appears to have been manipulated only in the counties using paper ballots. This makes sense because election officials and workers are much more likely to be able to grok ballot box stuffing and other such low-tech techniques.