Friday, June 5, 2009

Instant Runoff Virus errors found in Aspen vote totals

They incorrectly used Cambridge voting rules and software in Aspen. This is one of the main reasons why you can't successfully pull off voting on the cheap with IRV using uncertified software without making sure that all the rules are followed.

It took two weeks to detect these problems. But what about the fact that TrueBallot was allowed to use the wrong software for the job? How come no one in Aspen made sure the correct software was being used before hand?

They only had 2544 ballots to recount. If this happened in a larger county with over 60 times the number of ballots - like Wake County with over 150K ballots in the May 2008 primary election - it would have taken 60 times two weeks - or 120 weeks - or 2.5 years to do the same sort of audit.

But it is interesting that here in Wake County, they did the IRV "audit" which amounted to a non-public recounting of approximately 3000 ballots in one day.

That is why IRV is too complex to be practical for use in all but the smallest elections, where you can either spend two weeks trying to make things appear as though they worked out, or you can just do it in your locked office.

But even the smallest local races deserve just as much election integrity as larger races. Even more so, because our local races are the ones where we have the most potential to interact with our elected leaders. Yet these are precisely the elections most vulnerable to the "Instant Runoff Virus".
What about the Council races - including the one where they had two candidates left and no one crossed the threshold? Where did they pull the votes for the winning candidate from?

Chris Telesca

Last-minute note - Just got this from Marilyn Marks:

the “recount” was not an official recount, and it is uncertain as to whether it will be certified. Seems like a political stunt for the mayor.

But it was not a hand count. Done with the True Ballot digital data and software.

The data was not found (?) or disclosed until after the period for recount had expired.

So, more votes would not necessarily have made for a longer counting process.

Interesting that they went to the trouble to just sort through the data and software only after the period for the recount had expired. Had the race not been certified, someone could have challenged the results of the race and demanded a fill hand-to-eye recount of all the ballots.

That would have been fun - because unless you can set up a hand-sort that can do an accurate sort and the results can agree - you don't have a way to really audit the race.

Ireland wins Aspen mayoral race again
Error found in instant runoff tally, giving opponent Marks 16 fewer votes
Carolyn Sackariason
The Aspen Times
Aspen, CO Colorado,
ASPEN — Due to a computer software error, it turns out Mick Ireland won the Aspen mayoral seat on May 5 by a larger margin than originally reported.

City staff recently learned of the error in the tabulation of the final-round vote totals for mayor. However, the error had no effect on the outcome of the race, according to city officials.

Instead of Ireland defeating opponent Marilyn Marks by a vote of 1,273 to 1,140 (52.8 percent to 47.2 percent) as earlier reported, he actually won by a margin of 1,301 to 1,124 (53.6 percent to 46.4 percent).

In the mayor’s race, the threshold necessary to be sure of a victory was 1,273, which is 50 percent plus one of the 2,544 ballots cast. When Ireland reached 1,273 after LJ Erspamer was eliminated in the final round of counting, he was guaranteed to win.

The software stopped counting any additional votes for him. However, any ballots ranking Marks after Ireland were added to her final round totals.

Ireland ultimately received 28 votes beyond the threshold of 1,273. Sixteen of them had been counted for Marks, and 12 had been deemed “exhausted” because they did not rank Marks.

“City staff has been working to audit the instant runoff process, and the tabulation error was recently discovered by TrueBallot, the company hired by the city to perform the election,” said City Clerk Kathryn Koch. “The error arose because the voting software was originally written to support the ‘ranked choice’ form of elections used in Cambridge, Mass. Following Cambridge rules, the software prevented a candidate who had reached the winning threshold from receiving any more votes.”

Vote totals in all other rounds of the mayor’s instant runoff voting tally and in all rounds of the two council tallies were unchanged. The error did not occur in either of the council tallies.

“The fact that this error was detectable using election data we made available to the public validates our approach to election transparency and integrity,” Koch said.

Two days after the election, city officials and members of the public conducted an audit that involved randomly selecting 10 percent of the ballots and double-checking that the rankings corresponded to the electronic records.

The second step was to manually verify that every ranking was tallied correctly for mayor and council, which was conducted by TrueBallot.

Marks, who has been a critic of instant runoff voting well before it was implemented, said she thinks the testing of the system was inadequate leading up to the election.

“I am thrilled that there is some post-election auditing going on,” she said. “This demonstrated that it needed to be done.

“I hope that the results they’ve found will encourage them to do further work and tests.”

TrueBallot did the manual verification as part of their standard post-election services.

“We were able to audit and document this election more completely than any other public election that we have held,” Koch said, adding members of the public can view the data files that rank the candidates, as well as other election data public like TrueBallot’s 72-page spreadsheet that provides analysis of the results.

The election results are summarized on the city’s website at

The Aspen City Council has committed to reviewing instant runoff voting and its procedures at a public meeting to be scheduled sometime this summer.

Koch said she decided to make the error public by distributing a press release since the election data is available for public review. She added that to her knowledge, there are no other discrepancies with the instant runoff voting system or the election results.

“I’m confident we got the most correct answer,” she said.

No comments: