Ohio's Secretary of State Jennifer Brunner is planning "a first-of-its-kind audit of votes from the March 4 presidential primaries, saying the outcome should help ensure the integrity of future elections," according to a report this morning from the Columbus Dispatch.
"Brunner is calling on 11 counties to volunteer for the audit, in which at least 7 percent of the votes cast in each county would be rechecked by hand," the paper reports.
That's good. But there are a couple of points that we hope she is well aware of, since such post-election audits can offer a deceptive result in a number of cases...especially on touch-screen machines...and especially in Ohio...
The report from the Dispatch notes:
"During her campaign, Secretary of State Brunner talked about elections that could be audited or verified," spokesman Jeff Ortega said. "This is a further step toward that fulfillment."
A 2007 audit of Cuyahoga County's November 2006 election found a number of irregularities, including the loss of some ballots and others that were counted twice. Brunner praised that audit, which was ordered by the county board of elections, and suggested that other checks would be coming to ensure the accuracy of polling across Ohio.
That previously commissioned audit, of the Diebold touch-screens in Cuyahoga County after the '06 election, was very helpful. Among the things found were that some 10% of the "paper trails" on the systems were "destroyed, blank, illegible, missing, taped together or otherwise compromised." CNN's Lou Dobbs Tonight (video and transcript here) reported that the audit revealed on many of the touch-screens, "the machine's four sources of vote totals, individual ballots, paper trail summary, election archives, and the memory cards, did not all match up. The totals were all different."
The study helped lead, in part, to scrapping the touch-screen systems all together in Cuyahoga, Ohio's largest voting jurisdiction --- its largest city, Cleveland, was the site of enormous election problems in 2004 --- and the move to optical-scan paper ballots to be tallied at the county election headquarters.
But auditing "paper trails" from touch-screen systems can also be deceptive, and result in a false sense of security. A number of studies, including a landmark report from the NYU Brennan Center for Justice, revealed how such a system can be hacked to effect an election outcome, even while ensuring that the so-called "voter verifiable paper audit trails" (VVPAT) would still match up with the internal machine numbers in such a way that an audit would not likely discover the hack. Worse, such an audit --- of a hacked DRE/touch-screen election --- could actually result in a report claiming the election was "100% accurate," even though it had been gamed.
The take-away here: Audits of "paper trails" may be helpful in finding problems with touch-screen voting systems. Audits of such systems which don't find any problems when comparing "paper trail" results with machine results, may be exceedingly deceptive, and shouldn't be taken as a sign that the touch-screens worked correctly. There is simply no way to use touch-screen voting machines safely --- with or without a "paper trail" --- in any American election.
One other caveat concerning Brunner's proposed audits. As the Dispatch reports, "Local elections boards will randomly select precincts that account for at least 7 percent of the votes cast, and bipartisan monitoring teams will count ballots from the presidential primaries by hand."
The trick here is assuring randomness. After the 2004 Presidential Election in Ohio --- in Cuyahoga County, in fact --- Election Officials pre-counted the paper ballots in order to assure that their mandated 3% of precincts counted in the Green/Libertarian Party recounts would match the official reported results. The "random" selection of ballots was little more than a show, as the election workers rigged the recount. Two of them, were eventually sentenced to 18 months in jail for their felony crime.
The verdict was thrown out on a technicality, and before a retrial could occur, the two women convicted pleaded "no contest" and were given no time in jail.
The take-away here: Make sure "random" really means random!
[UPDATE: EmLev points out in comments that there may indeed be a problem with Brunner's definition of "random"]
Good luck, Ohio!