Pierce County Washington adopted "instant runoff voting" in November 2006. Pierce County officials were backed into a corner in May 2008 and ended up getting permission to use uncertified IRV software, violating their own state's laws. Officials said that even counting just 14 IRV ballots was too hard, and they needed the state to approve the uncertified software. Their machines could not tabulate instant runoff so they were waiting for the vendor to develop software. Time was running out because Pierce was supposed to utilize Instant runoff for their upcoming elections. The vendor delivered the software but it hadn't been federally tested or approved at the time.
Pierce County tested the new Sequoia software, found some problems, and asked the Secretary of State to certify parts of the system on an emergency basis since it would be too hard to count manually.
Pierce County officials said: they tried hand-counting just 14 RCV ballots with seven ranked contests and found that it was “horrendous.” Using software to tally this sort of balloting was absolutely essential. She found that it simply couldn’t be done any other way.
Ellen Theisen and John Gideon of Voters Unite have a writeup on the May 13 testing results and for the IRV software being considered for Pierce County voting machines. found serious problems with them.
In testing, the system reported that there were "zero votes" to start(in one set of its records), but actually had 56 votes already in its other set of vote records.
Voting System Certification Hearing in Washington State May 23, 2008 by Ellen Theisen
...John and I were concerned about the Sequoia system. Pierce County, WA voters had voted to use Ranked Choice Voting (RCV), a system where voters may specify their preferences in particular races on the ballot as "first choice", "second choice", or "third choice". However, no state-certified system had the software needed to tabulate RCV ballots. So Sequoia developed their software specifically for Pierce County. They had applied for “emergency” provisional certification from the state, which would allow them to bypass state requirements for independent testing to the federal standards.
John and I planned to go to the testing on May 14, but at the last minute the testing for that day was called off so officials could investigate a problem that had shown up the previous day. the Sequoia system had tallied votes wrong.
Later in the day, Patty Murphy emailed us from the SoS office to let us know what happened:
Monday night, a test run was done on the 5 insights [Sequoia Insight Optical Scanners]. Once complete, one tester zeroed out each machine and memory pack. (You can reset, or zero out, the memory pack at the Insight machine.) Following this tester was a 2nd worker who powered off all the Insights. On one of the machines, they powered it off before the machine was finished zeroing out the memory pack. Each memory pack has all the results tabulated by precinct first on the pack, followed by cast vote records that have results by ballot. It zeroed out the tabulated section, but it didn't erase all the cast vote records.
On Tuesday, when we started the machines, we ran zero reports. They were all zero because they show all the tabulated results by precinct (the portion that was zeroed out). However, when we moved the memory packs to the reporting software, it added up all the cast vote records, and reported those results.
In other words, the Sequoia system checked one set of electronic ballot records and reported an “empty ballot box,” but then used a different set of ballot records to tabulate the votes. This is exactly like checking one ballot box to ensure that it’s empty before opening the polls, and then using a different, unchecked ballot box for the voted ballots to be tallied at the end of the day. But the unchecked box hadn’t been empty.
John and I both wrote up testimony, which we immediately sent to Patty Murphy. As you can imagine, we objected to having a system certified when, in testing, it reported that there were "zero votes" to start, but actually had 56 votes already in its secret, unobservable "ballot box"! Patty Murphy passed our testimony along to the Review Board so they would have time to read our statements before the hearing.
Here's John's. Here's mine.
Patty thoroughly described the problem that had occurred, the discrepancy in the results, and the possible workarounds proposed by Sequoia —which involved “patching” the software in both the scanner that falsely reported a blank ballot box and in the WinEDS software that counted ballots from the “stuffed” ballot box.
Then she added a new thought we hadn’t heard before — the possibility of using the Edge touch-screen machine, the central count optical scanner, and the WinEDS tabulating software without using the flawed Insight scanner. I thought that sounded like a possible solution.
Regarding the complex algorithm used to tally RCV ballots, Patty pointed out that software, of some type, would be needed to tally such ballots correctly. She explained how they tried hand-counting just 14 RCV ballots with seven ranked contests and found that it was “horrendous.” Using software to tally this sort of balloting was absolutely essential. She found that it simply couldn’t be done any other way.
...Pat McCarthy, Pierce County Auditor, spoke first. She said that she was so concerned about the flaw on the Insight scanner, she had decided not to use it at all, contrary to her original intention. ...Pat McCarthy had listened to the evidence and had
decided to reject the equipment, regardless of what the Board recommended or whether the SoS certified it.
...Edwin Smith from Sequoia was there to represent their system that had failed during testing. He attempted to claim that the miscount wasn’t really an accuracy issue, because it had been caused by an operator error that could be handled through proper procedures. But the Review Board heard John’s statements about the accuracy requirements in federal and state law and acted on them responsibly -- as you’ll read in a moment.
I pointed out that what Smith called “operator error” was really good testing, and I saw heads nod among the Review Board.
...To get us back on track, John emphasized again that using a system that miscounted votes was a violation of federal law that could not be ignored.
Debbie Cook ...and made a motion to recommend approval for the system, with a stipulation prohibiting the use of the Insight scanner. Joachin Avila suggested adding another amendment — the approval would depend on a legal decision that the system would not violate the accuracy requirement of federal law as long as the Insight wasn’t used.
The Review Board unanimously passed the motion as amended.
The Secretary of State of Washington granted "emergency" permission in May 2008 for Pierce County to use the uncertified software on Seqouia machines, even though flaws were found in the WinEDS (central tabulating system). The "Insight" optical scanners for precinct voting were not approved, but the touchscreens and central count scanners were - on an emergency basis. All optical scan ballots will be hauled off to the county office to be tabulated.
(June 13 2008 Secretary of State press release)
This report has been forwarded to the San Francisco Elections Department, since they plan to use the Sequoia "Insight" precinct scanner once federal certification is granted. iF the machines are not certified by November, then San Francisco will not use them to tabulate the IRV rankings. That will be done by hand.