Florida needs to demonstrate that it is a functioning democracy by getting to the bottom of the problem in the District 13 Congressional Race in Sarasota County. Republican Vern Buchanan now leads Christine Jennings by only 369 votes. But voter complaints and anomalous statistics raise the question: did the voters decide this election, or did software or hardware errors tip the result in favor of a candidate who would otherwise have lost?
The numbers reported the day after the election showed a glaring problem: over 18,000 voters cast ballots, but (supposedly) did not vote in the Congressional race. Overall, that is a 14% undervote rate, which is implausibly high for this race (the undervote rate in the U.S. Senate race was a little over 1%).
According to Supervisor of Elections Kathy Dent and candidate
Buchanan, the undervotes were protests by voters repelled by a
negative campaign. This argument does not pass the straight face
test. To believe it, one would have to accept that the only voters
who were unhappy were those who voted on electronic machines in
Sarasota County. In contrast, the undervote rate on Sarasota's paper
absentee ballots was about 2.5%. The undervote was similarly low in
neighboring counties.
From what we know now, the voters probably elected Jennings, but the
voting machines elected Buchanan. Even with the lost votes, Jennings
got the most votes in Sarasota County (neighboring counties preferred
Buchanan, which is why he's ahead now). If the 18,382 lost votes are
distributed in the same proportion between the two candidates as the
votes that were not lost, Jennings would win by about 600 votes.
What was the problem with the machines? Some are quoted in the press
saying that a confusing display caused voters to miss the
Congressional race. However, there were equally bad display problems
with races further down the ballot that had lower undervote rates.
Furthermore, many voters swear that they definitely selected a
Congressional candidate, but the vote disappeared later when a summary
screen was displayed. Therefore, it is a good bet that there is a
more serious problem with the machines.
Instead of speculating, we need a serious investigation to find out
the truth. In my opinion, there are three important avenues for
investigation:
A DETAILED ANALYSIS OF THE ELECTRONIC AUDIT LOGS IN THE MACHINES.
This could tell us whether the machines were recording error conditions,
whether the errors were concentrated on particular machines, and
whether one candidate or the other lost more votes.
EXTENSIVE TEST VOTING ON A VARIETY OF MACHINES BY A VARIETY OF VOTERS.
If voters experienced problems, test voters would probably also
experience them. However, the problems could be specific to
particular machines, or the behavior of particular voters, hence the
need to test with various machines and voters. Analysis of the
machine-by-machine results could show that certain machines are more
error-prone than others; of course, those machines should be tested
specifically.
PRECISION TESTING OF TOUCH SCREENS.
The touch screen and the display
on these machines are separate devices, and the computer needs to keep
them aligned. Some of the problems reported by voters could be
explained if the area on the touch screen that registers a touch over
a "button" displayed on the screen were misaligned, too small, or the
wrong shape. That could be determined by touching the screen with a
sharp probe to map out the shape of the areas corresponding to
displayed buttons.
Initial findings must be followed up by further investigation. The
manufacturer must be compelled to cooperate with independent experts
to examine the machine hardware and software to make sure that the
problem has been fully explained.
Sadly, almost two weeks after the election, these tests have not even
been started. The Florida Secretary of State's office did map out a
multi-page plan for an "audit" that called for test voting and
examination of a few audit logs. This audit would be an incomplete
and inefficient investigation, but it would deal with some of the
basic issues. Absurdly, those tests have been delayed because the
Buchanan didn't have an expert to observe them!
The election will be probably be certified on Monday. After that, the
next step would be a court challenge. If that happens, the court
should insist on a thorough and transparent technical investigation,
including, at a minimum, the tests suggested above. If it is
determined that the result of the race was decided by a machine error,
the only fair resolution is a re-vote in Sarasota for that race
(preferably on paper ballots).
This problem was inevitable. In 2004, complaints about voting
machines changing or losing votes were lodged by voters all over the
country, on several types of electronic voting machines. The problem
was obvious then, and I called for an investigation before the next
election. Nothing happened, and there were widespread complaints
again this year. The complaints on other equipment and in other parts
of the country need to be investigated, urgently, or machine problems
will lead to more disputed elections in the future.