Sunday, October 14, 2012

Untied Alternatives

C. G. P. Grey has another election-themed video out and, like the last one, I have one tiny nitpick from the very end of the video. First, the video:

Right near the end, Grey suggests using a preferential vote as a way to avoid the ridiculous series of ties and tie-breakers he discusses, and by preferential vote I presume (since he made a video arguing in favor of it) the "alternative vote," AKA, instant runoff voting.

The problem I have is that IRV is actually more likely to run into problems with ties than most other voting systems; including plurality. This is because each elimination done under IRV, even the ones far down the list who received very few first place votes, could potentially be a tie, and effect the final winner. As always, a quick example:

  • 45%: A > (others)
  • 25%: B > C > (others)
  • 15%: C > A > (others)
  • 15%: D > C > (others)

Candidates A and B lead, but candidates C and D tie for third place. But how you break that tie determines who wins the election. If you eliminate C first, then A wins, while if you eliminate D first, then C wins. And it can get much worse; if A and B were instead at 40% and 30%, then after the first tie-breaker, we could have had another tie. Since each tie—or near tie— result can require a recount and potentially a series of lawsuits, IRV has the potential to be a total nightmare as the elimination of each minimally-supported candidates is fought out in court (supported by whichever candidate up the chain benefits.)

This is all, of course, extremely unlikely. A near-tie in a Presidential election has only gone to the Supreme Court for review once (in 2000) and that hopefully won't happen again anytime soon. But we can say, thanks to computer simulation, that IRV elections are about 1.8 times more likely to have a results-effecting tie than plurality elections, even with just three candidates.

So yes, the electoral college is kind of silly, and the tie-breakers for it even more so. But IRV wouldn't help with that.


  1. It would if it were modified as I advocate it to be modified.

    Why are you flogging old IRV when there's a better model on the table that FairVote will hopefully adopt...
    (It's under consideration I believe.. and doesn't require recursion to explain...)

    If there were a statistical tie in the first stage for which candidate received the third highest amount of ranked votes then one could do the runoff for two sets of three candidates and see if it affects the winner. The odds are it wouldn't.


  2. My apologies, I didn't read your post thoroughly yesterday. Arguably, if IRV forces the two front-runners to move closer to the center then yes it ought to lead to more ties than FPTP.

    My thoughts though are that Ken Arrow is right that the score-voting would not be included simply out of prejudice towards models based on theoretical assumptions.

    But if one wanted to prevent an ugly tie, a non-instant runoff might work.... Like my idea to determine 3 winners in a General Open Primary and then let the Electoral College determine the final winner within a week. If there was a statistical tie for 3rd place then 4 candidates could advance to the EC, where they would vote until one candidate got a majority!