By Aaron Blake - 01/24/07 12:00 AM EST
A new academic study of voting in Florida’s 13th district congressional race refutes the prevailing opinion of experts that the large undervote in the race in November was likely due to poor ballot design.
Democrat Christine Jennings lost by 369 votes out of 240,000 cast, and has contended that the undervote — more than 18,000 Sarasota County ballots didn’t include a vote in the congressional race — could be attributed to malfunctioning electronic voting machines.
The new study points to an error message that the experts say correlates with the undervotes, as well as to voting patterns that show unusually high undervote rates for Democratic voters. The study’s authors say more information, specifically the secret software used by the voting machines, is required to draw any real conclusions.
The Jennings campaign plans to tout the study, which was researched independently by Stanford University computer science professor David Dill and Cornell University government professor Walter Mebane, as support for its appeal in a Florida court.
Jennings is seeking access to the source codes for the voting machines in hopes of proving the machines malfunctioned and, ultimately, gaining a new election. Poor ballot design is not sufficient reason to redo the election.
“This study provides yet more evidence about why a thorough review by outside experts is the only way to end the crisis in confidence in District 13 and among millions of America’s voters,” Jennings’s attorney, Kendall Coffey, said.
The declared winner in the race, Rep. Vern Buchanan (R-Fla.), was seated provisionally earlier this month, meaning the House reserves the right to remove him pending the case in Florida. The House Administration Committee also has taken up the matter.
Late last month, a Florida judge denied Jennings’s request for access to the source codes. She is appealing the decision.
In the new study, Dill and Mebane contend that while ballot design likely contributed to the undervote, it doesn’t completely explain it. They are not working with the Jennings campaign, but have worked with two nonpartisan groups that also are contesting the results.
“Whether [Jennings is] going to use or abuse this report, I do not know,” Dill said. “But our point is that further investigation is warranted.”
The study’s authors’ main argument focuses on the event log files, which report the machines’ transactions. They say there is a “very strong statistical link between a specific error message in the machine’s event log and a high undervote rate on that machine.” This, they say, suggests a mechanical reason for the undervote.
The study found voters were significantly more likely to register undervotes in the congressional race if they voted Democratic in the Commissioner of Agriculture race, regardless of other votes. Also, those who voted straight-Democratic in the five statewide races were more likely to cast undervotes than straight-Republican voters.
“When I found the correlation between the undervote rate and the error message, my hair stood on end, and I said, ‘Whoa, that’s something’ — and then even more when we found the interaction between the undervote and the voter preferences,” Mebane said. “That, even more, suggested to me that something really problematic is going on that needs to be looked at.”
Mebane said the research has reached the limits of what statistical analysis can provide and, with nothing proven at this point, the software should be made available.
He said he would also urge examination of the “Personalized Electronic Ballots,” or PEBs, little devices that sit atop voting machines and communicate with them. The error message Dill and Mebane cite, which appears on the event logs as “Invalid vote PEB,” suggests a malfunction could have something to do with the PEBs, they say.
Buchanan’s lawyer, Hayden Dempsey, emphasized that the study contained no conclusive evidence that the machines caused any problem and that it also confirmed that a number of other explanations for the undervote were plausible.
“Under Florida law, the judge has ruled that their case is meritless and based upon speculation and conjecture and that they’re not entitled to see the source code,” Dempsey said. “In the meanwhile, test after test and recounts keep showing that the machines worked as they were supposed to.”
Other experts have said the undervote was likely the result of poor ballot design, as the two-candidate congressional race was paired on a screen with the dominant gubernatorial race, which featured many candidates.
They point to other high undervotes for a different race in nearby counties, where the two-candidate state attorney general race was paired with a race that dominated the screen. They say voters probably missed the races inadvertently.
Dill and Mebane cite another similar screen on the Sarasota ballots on which a Hospital Board race could feasibly have been missed by voters. But those who undervoted in the congressional race undervoted much less often in the Hospital Board race, and it’s unclear why whatever made them miss the congressional race did not cause them to miss the Hospital Board race, Dill and Mebane argue.
Jennings, who is in Washington raising money, meeting with lawmakers and attending the State of the Union address, said she remains confident that her efforts will pay off. She also has statements from Sarasota County voters about problems they encountered while voting.
“I think that this election is not over,” Jennings said. “And I think our evidence is so compelling for a new election.”