Aspen CO 2009: IRV mayoral & city council STV elections

Not exactly an inspiring success either for STV, or for accurate vote-counting

Rob Richie, the director of FairVote, a Washington DC based IRV-advocacy group, traveled to Aspen to testify before the town council and urge Aspen to switch to an "instant runoff" (IRV) election system. Another IRV-advocate who did the same was Caleb Kleppner, one of the founders of TrueBallot Inc, a company selling IRV vote-counting software. Kleppner also was listed as a FairVote "senior advisor." It worked. A Nov. 2007 referendum decided on IRV by 72%. Unfortunately only 837 voters participated in this referendum (16% turnout) so this was not as decisive a victory as it might have seemed.

The first IRV elections duly occurred on 5 May 2009 (for mayor) and at the same time the city council was elected using (rather peculiar nonproportional) STV rules. The reason for the peculiar nature of these STV rules was that the new system was intended to simulate Aspen's old runoff system, but "instantly." (The ordinance [pdf].)

As usual, the IRV propagandists at "Fair"Vote immediately blared their trumpets about how this election had been a "great success," just like they had after the Burlington 2009 mayoral election disaster a few months before. As far as I've been able to tell, the concept that an IRV election might not be a great success does not exist in FairVote's universe, so they apparently pre-write their "it was a great success" press releases before each IRV election so they can rapidly release them ASAP.

But unfortunately, Aspen's elections had some serious problems conveniently left unmentioned by the IRV propagandists.

  1. Mick Ireland was elected mayor with exactly 1273 votes as his official total. Also elected with exactly 1273 votes was city councillor Derek Johnson. Also elected with exactly 1273 votes was city councillor Torre. (Incidentally, Torre is one of those nowadays-rare people who goes by only one name.) Amazingly enough, in all three cases 1273 happened to be exactly 50% plus 1 voters. Coincidence? These are still the "official" results as of 13 June says
  2. This was lauded by FairVote chief executive Rob Richie as: Also notable were the fundraising figures. Challenger Marilyn Marks outspent Ireland, breaking Aspen records with almost $40,000 in funds. Ireland mustered less than half of Marks' total, with less than $18,000 raised. Despite this disparity in resources, Ireland emerged victorious.
  3. However, after this "coincidence" was pointed out to her, Aspen city clerk Kathryn Koch felt differently. She then wrote a press release, dated 28 May, saying TrueBallot's results had been erroneous, and, e.g, Ireland really should have gotten 28 more votes. According to that press release, the problem was partly that the election was counted by software from TrueBallot Inc. (This same company's software had also been used to count the votes in Burlington's mayoral election.) This software had been set to stop counting votes as soon as any candidate acquired 50% plus one votes! Evidently, in the real world, Aspen and/or TrueBallot Inc. had been unable to perceive that this might not be the best possible way to employ this software... The other part of the problem was that some votes had been counted erroneously, specifically 16 Ireland votes had been erroneously counted for his opponent Marilyn Marks. As a result the Ireland-Marks margin, originally announced as 1273-1140, was changed to 1301-1124. Ireland still won, though.
  4. Almost certainly one or both of the other two 1273-vote "winning counts" were also erroneous for the same reasons, but as of this writing they have not been officially corrected. Harvie Branscomb's analysis below agreed that Torre and D.Johnson indeed won, but their true totals were 1074 for Torre and 1234 for DJ. The reason the TrueBallot software (apparently?) gave them both 1273 was it wrongly thought that 1273 was the official margin, since 1273 is 50% plus 1 of the 2544 ballots cast. But actually, according to Branscomb, "2509 ballots were voted for either 1st or 2nd preference or both in the city council contest, therefore according to the rules the initial threshold for round one is 1255." Apparently the TrueBallot software ignored the rules and ignored the totals, and then because it had (apparently correctly) found Torre and DJ to be the winners, it gave them both a pretend-number of 1273 votes, and declared them victorious. This also had the enjoyable-sounding side effect of causing them to be "majority winners" even though in fact, with 1234 and 1074 votes they both were sub-majority winners.
  5. After the (actual) votes had been released by Aspen, they were analysed by Harvie Branscomb (and also separately by Marilyn Marks' nephew Douglas Marks, who found the same results). Branscomb's analysis (pdf) showed that the STV council election exhibited non-monotonicity. Specifically: If 71 Michael Behrendt voters had instead voted Jack Johnson top, above Behrendt (leaving their ballots otherwise unaltered), that would have made Behrendt win!
    (In the actual election Behrendt did not win, winners were Derek Johnson and Torre. Of course, the precise number "71" is not essential, this is merely the fewest such voters needed to make Behrendt win in this insane-seeming manner. Any value from 71 up to 78 works.)
    Behrendt thus could have stated the following legitimate gripe:
    "I would have won even with the handicap of shifting 75 of my voters to Jack Johnson. So why the heck didn't I win without that handicap?"
    Similarly his voters might be rather distressed to learn that by voting for Behrendt, they caused him to lose, and really (if they hadn't been strategic idiots) they should have voted Jack Johnson to make Behrendt win!
  6. Marilyn Marks (who came in second to Mick Ireland in the Mayoral race) wrote this open letter to FairVote head Rob Richie, describing numerous real-world problems Aspen experienced trying to run its elections using IRV.
  7. The Aspen Times conducted a poll asking, in hindsight, "Should the city of Aspen continue using instant runoff voting for City Council elections?" The results on 6 June 2009 (exactly 1 month after the election) were 99 NO, 75 YES, and 15 UNDECIDED out of 189 total. Based on this, it would appear that Aspen believes its earlier decision to adopt IRV was a mistake. Marilyn Marks also conducted an online poll. Her results as of 6 June also indicated that Aspen residents now were considerably less enthusiastic about IRV in hindsight than they had been back in 2007:
    1. Did you support Instant Runoff Voting when it was proposed in 2007? YES 21, NO 33, OTHER 23.
    2. Did you vote in the Aspen 2009 election (which used IRV)? YES 53, NO 1, OTHER 23.
    3. Do you believe that the results of the Council race would have changed if there had been a traditional run off among Derek Johnson, Jack Johnson, Michael Behrendt, Torre, and Adam Frisch? YES 49, NO 8, OTHER 20.
    4. Would you prefer to continue to use the IRV system in future elections? YES 23, NO 52, OTHER 18.

  8. Later note: Aspen has now repealed this system.

Return to main page