Election Forensics: What They Can -- and Can't -- Tell Us
November 08, 2004
Posted by Donna Wentworth at 10:56 AM
DEEPLINKS Blog Entry From: http://www.eff.org/deeplinks/archives/002073.php
The Internet is buzzing about whether the electronic voting systems used in this election really worked as "well" as they appeared to work. Is it
possible that some machines malfunctioned in ways that skewed results? Could the 4,530 votes lost due to a data storage error be only the tip of the e-voting iceberg?
The good news first: From what we can tell, it is unlikely that the problems with touchscreen machines changed the outcome of the
presidential race. But that doesn't make it impossible, and EFF is still looking into some problems in Ohio and
elsewhere that could be very important.
The bad news: Let's suppose for a moment that the picture of the presidential race stays unchanged. Does this mean, as some vendors are claiming, that the machines "passed the test"? Nope. If the election had been closer in such key states as Florida,
Pennsylvania, New Mexico, or even Ohio, the problems we saw could easily have thrown this election into chaos, and that
chaos could have affected either candidate.
It will take some time to analyze the information recorded through the Election Incident Reporting System (EIRS), but regardless of what we find, the current figures show that machine malfunctions were the third most common voting
problem reported. And recent reports show that not all problems were obvious.
EFF will be moving in the next few days to seek to examine the machines exhibiting the most troubling malfunctions. The
goal is to figure out whether what we saw indicates even more serious problems.
Which brings us, finally, to the ugly news: There's one story about this election that we'll never know: what happened inside the machines that do not have a
paper trail. It's somewhat reassuring that, in most instances at least, final exit polls and other external systems give
us roughly the same picture that the election results do. But suppose that wasn't the case? This is what audit trails
are for. The figures in cooked books often look perfectly fine; so would a cooked vote tally. In this election, we are
forced to take it on faith that our votes were recorded in the way that we intended. But as the late former President
Reagan noted long ago, when important issues are at stake, we need to both "trust" and "verify." That's why the battle continues to persuade election officials nationwide to adopt systems that are 1.) verified by the voter, and 2.) can be
audited after the fact. We can't let bullish pronouncements by vendors persuade people that an election that cannot be
audited "works." This was an important election, but it's not the only one.
On Friday, three Democratic Congressmen sent a letter to the General Accounting Office requesting an investigation into voting irregularities caused by e-voting machines.
This isn't a bid for a recount; they do not anticipate that an investigation will turn up anything that would change the
election results. It's about securing the future of the democratic process for everyone. Let's all hope they get a warm
reception.
For more information about the malfunctions reported on Election Day and EFF's concerns, listen to our press conference,
available here [MP3 format] and here [OGG format].
Sequoia Acknowledges E-voting Problems
November 09, 2004
One of the most serious problems with touchscreen voting machines reported on Election Day was the misrecording of
votes, sometimes called "ghost voting" or "jumping votes." This is when a voter attempts to vote for one candidate and
the machine indicates that he or she has voted for another, either immediately or at the summary screen stage. In an
even more disturbing variation of the problem, some voters discovered that the machine had "pre-voted" for a candidate
or even a slate of candidates, and that they consequently had to "unvote" the default choices.
Despite the anxiety and disruption this caused, voting machine vendors preemptively declared that touchscreen machines passed the test, suggesting that any problems should be attributed to "human error." Now,
perhaps recognizing that discussion of the problem isn't going away, one leading vendor, Sequoia, has finally admitted that the problem exists. But, true to form, the company is evidently blaming poll workers for Election Day mistakes.
Yet the responsibility for providing tools that poll workers can use properly given the limited amount of training they
receive lies with the vendors and voting officials.
As we noted yesterday, EFF is investigating the most troubling e-voting incidents reported through the Election Incident
Reporting System (EIRS) to determine whether they indicate serious machine malfunctions. We invite every voter who
experienced "vote jumping" to contact us to share specific details about your experience. You may help us establish that the pattern we're seeing is something
that vendors and voting officials can and must address.
ENDS