Put bluntly, MIT's chosen tool --risk-limiting audits -- cannot expeditiously cross the finish line in photo-finish elections or in recounts in high-population jurisdictions.
"It is a solution that is not practicable, particularly for the large practitioner," said Ion Sancho, who recently retired as the Supervisor of Elections in Leon County, Florida, after nearly three decades in that job, and who helped develop the ballot-image audit system just endorsed by his former peers. "It [an RLA] is scientifically valid, what they are doing. But it doesn't lend itself to the kind of organized, pre-planned activity that matches everything else we're doing. Everything has to be pre-planned. The process has to be as efficient as possible... How do you plan [an RLA] for what happened in Florida with three races within one-half of 1 percent, and two within one-quarter of 1 percent?"
These pros and cons do not mean that this array of new audit tools should not be used. These technologies and techniques should be deployed -- but deployed knowingly and judiciously. They should be targeted to the task at hand -- not oversold. They should be assessed to see how they could be integrated into a more streamlined ballot inventory, vote-verifying and confidence-building process.
For example, RLAs require that all of the ballots in a race to be audited be assembled in one location. In Colorado, which votes entirely by mail, that's more easily done than in states using precincts, mail and vote centers. Once assembled, the numbers of ballots in storage boxes must be checked before an RLA can start pulling random ballots to physically examine and start its math. Contrast that process and ballot inventory with what Florida officials seek to do: a second scan (no data from the initial tabulation is used -- only the actual ballots) to create a new tally and searchable library of every vote, and where the ballots are scanned in batches tied to the first scanners used (to trace machine errors).
"Here's the difference as a practitioner of elections: good enough is not good enough for me," said Sancho. "I've dealt with this issue for such a long time. It's one of the reasons I pulled back from the Electronic Verification Network [including RLA co-authors]. I really like those individuals... but I want to know what I can do to train out of my citizenry the typical mistakes they are making, which may make their ballot or their vote not count, or slow down the process, or reduce the overall efficiency."
Stepping back, this spectrum of emerging audit tools and a push for verifying votes is a positive development. It is a counterpoint to post-truth rhetoric shadowing political campaigns. But the factions inside the insular world of verifying votes should heed Matt Masterson's cautionary note -- there is no one-size-fits-all solution. They should ponder Ned Foley's historic frame: how accurate do Americans expect their vote counts to be? And officials buying new voting systems should ask themselves if they will be able to withstand the propagandistic attacks that Walter Mebane said will be forthcoming.
Why can't paper ballot libraries be built that support RLAs and ballot image audits before certifying outcomes? Why can't random ballot selection techniques affirm that digital ballot images are correct? Why can't hand counts -- which are less accurate than scans, according to new research based on the Green Party's 2016 presidential recount -- be selectively used after mountains of paper are turned into libraries? Why can't problematic ballots quickly be identified and pulled for scrutiny as the public watches online?
As Ned Foley said, how accurate do Americans want their vote counting to be?
This article was produced by Voting Booth, a project of the Independent Media Institute.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
