You all know that I don't like Instant Runoff Voting as a replacement for primary elections or runoff elections. Here's how Ranked Choice Voting is causing more problems in San Francisco, CA where they have used it for years!
Vote recounts costly, a roll of the dice
Marisa Lagos Updated 10:32 p.m., Saturday, December 1, 2012
Election recounts are an expensive, lengthy proposition in any race, but the close finish last month in a San Francisco supervisor's contest has raised the question of whether they would ever be affordable - or feasible - under the city's ranked-choice voting system.
Supporters of F.X. Crowley, who lost the District Seven supervisor's race to Norman Yee by just 132 votes, have taken the first steps to initiate a recount, but aren't sure if they will go through with it because of the cost. Under California law, the person who requests a recount must pay for it - and they don't get reimbursed unless the results change. In a two-candidate race, it usually costs around $1 a vote - but under ranked-choice voting, it could be more than double that amount.
It doesn't appear that a recount has ever been done under the system instituted by San Francisco eight years ago, which allows an unlimited number of people to run for office on a single ballot and is also used in Oakland, San Leandro, Minneapolis and several other cities.
While ranked-choice advocates insist that the difference in cost would be modest, a number of elections experts said that system would probably make a recount more complicated and longer, which could significantly increase the cost. The Department of Elections estimated Friday that a manual recount of the 35,140 ballots cast in the District Seven race would cost nearly $80,000, or more than $2 a vote. A machine recount would cost nearly as much.
That could easily put a recount out of reach for most San Francisco candidates. The biggest fundraisers in this year's supervisor's races, for example, brought in an average of $120,000 each, and under San Francisco law, candidates are only allowed to raise an additional $100 from each of their previous donors for a recount. That means it would almost always take the support of outside groups - such as the labor unions that have stepped into the Crowley race - to finance a recount, and even then it could be out of reach for many organizations.
In a normal race with just two candidates, those asking for the recount usually test the waters by first having elections officials pull and tally the ballots from a handful of precincts, to see if the results change. If they do, donors may be willing to contribute more money and continue the count.
But under the ranked-choice system, said San Francisco-based elections lawyer Randy Riddle, it may not make sense to incrementally recount votes by precinct.
Ranked choice allows voters to name their first, second and third choices; when the votes are tallied, the candidates with the fewest first-place votes are eliminated, and their supporters second and third choices are reassigned to the remaining candidates until one candidate receives more than 50 percent of the vote. In the District Seven race, there were six rounds of reassignments.
And that reassignment of votes is what complicates things, said Riddle, who is not involved in the Crowley recount.
'All or nothing'
"You need to look at the votes from the whole district to see how they are distributed," said Riddle. "It may have to be all or nothing ... and that could be expensive."
John Arntz, the city's elections chief, said how the agency retallies votes is up to those requesting the recount - but that the most likely scenario, a manual recount, "takes a lot of time."
Even using the regular tallying process - in which machines are used to count the majority of ballots - the District Seven race took more than a week to sort out, which is not unusual under ranked-choice voting.
Steven Hill, a consultant who helped draft the ranked-choice voting systems in San Francisco and Oakland, said he does not believe a recount is that much more complicated under the ranked-choice system. He and officials at the nonpartisan FairVote group said the bulk of any recount costs are associated with "the initial sorting and tallying of ballots that would be required of any election."
A preliminary step
In District Seven, Hill suggested that as a preliminary step - to gauge the likelihood of the results changing - elections officials could only recount the ballots where someone besides Yee or Crowley was in first place.
But Arntz said the results from that sort of tallying would not be legally binding. Additionally, said Jim Sutton, the lawyer representing the labor groups behind the Crowley effort, you have to recount every vote because there's no way of knowing how an individual ballot was counted the first time around.
"You're not comparing vote to vote because you have no idea which ballots are switching," he said. "Let's say a couple of first-place votes changed - that might also change the order of which candidate was thrown out first."
Initially, Sutton had believed that the best way for Crowley's backers to proceed was to have elections officials run all of the ballots through the voting machines a second time, before embarking on a much more expensive manual recount. But the campaign was weighing its options over the weekend after seeing the elections department's cost estimates that put a machine recount at just $7,000 less than a full, manual recount.
Machines aren't perfect
Sutton, who has been involved in several recounts around the state, said machines aren't perfect and that the most common problem is people marking ballots in a way that makes the machines reject their vote.
"People think a recount is about challenging (the validity) of ballots, but that's a very small part. People mark their ballots in such crazy ways, if I hadn't seen it I wouldn't believe it," he said.
Still, any recount is a gamble for donors.
"Statisticians will say it's probable that 132 votes will change in full manual recount - what's unknown is which way they will go," Sutton said. "That's the wild card."
Marisa Lagos is a San Francisco Chronicle staff writer. E-mail: email@example.com
We did go through some recounts of IRV races in NC. The 2007 Cary IRV race was recounted in secret - a few extra votes were found and a calculator error was discovered (pointed out by Don Hyatt and myself during the original IRV sort/stack/tally).
In the 2010 statewide Appellate Court race, a new wrinkle was added: undertested and uncertified IRV counting methods for DRE touchscreen votes (kind of against the original legislative intent of our "Public Confidence In Elections Act", or so Verla Insko and Ellie Kinnaird tell me).
Why does that matter? While fewer than 25% of NC counties use touchscreens, around 40% of the total statewide votes are cast on them.
The cost of doing a hand-to-eye recount of all the toilet-paper ballots from all the machines was so great that all the election directors in DRE counties got their county BOE members to write the SBOE and tell them they couldn't afford to follow the law and do it right. So the SBOE allowed the counties to port the voting data over to an undertested and uncertified method of a PC with MS Excel. A recount closed the gap between the two candidates.
But it did nothing to alter the strange fact that in all of IRV land, the first-row winner usually wins in the end (after counting all the 2nd and third column votes under whatever IRV scheme you use) about 95% of the time. In both of the 2010 judicial races that were decided by IRV - the second-place finisher in the first round flipped and won in the end.
But since Thigpen was taking public financing, he could not by law raise outside money to pay for a full forensic recount. So in my book, all three of the elections decided by IRV ought to have an asterisks by them. And it's all the more reason why we need to dump IRV and either change the law to lower the threshold or just allow for plurality winners. I mean - judicial IRV was pushed because the 2004 NC Supreme Court race was won with around 24% of the vote. The 2010 statewide judicial IRV race was won by slightly less than 28% - only a slightly larger plurality.