Rank with cutoff runoff
-
What do you think of this? A ranked ballot is submitted with an approval cutoff. The top two approved candidates are put head-to-head in a majoritarian runoff based on the rankings.
Probably choosing where to put the cutoff is not obvious and becomes intrinsically strategic, and related to the position of the two suspected front-runners in the rank. Maybe you would approve of every candidate you prefer to both suspected front-runners. I think it gets too complicated.
-
I think it a vast improvement over IRV Hare and choose-one plurality, and very similar to STAR.
I hope equal ranking would be permitted at the top if it is permitted at the bottom.
As a voter, I would rank by my true valuation. I would think of my best continuous Score vote and use that to determine probabilities for a stochastic process of setting the approval cutoff. This means that in, for example, a Bush/Nader/Gore contest, there would be a high probability of my falsely expressing approval for Gore.
-
@jack-waugh yes, you may be right. That strategy is close to the optimal strategy for approval. Except, if you believe Bush and Gore are going to be front-runners independent of your own approval, you don’t actually need to contribute to their approval scores if you don’t want to. Your ranking of them will still indicate your preference between them in the runoff even if you don’t formally approve them.
You could still give Nader the best chance of being a front-runner with the ballot N|>G>B. I think the biggest issue here is independence of clones. It is very similar to STAR. It basically is STAR with approval scoring, except you can indicate rankings between equally-scored candidates.
Here is a wacky alteration:
The top 2 approved candidates are identified. The runoff between them is not for election, but for elimination, I.e. the front-runner who loses the ranking runoff is removed from the election. The process is repeated over the remaining candidates until a single candidate remains.I think this addresses independence of clones but probably opens a can of worms with other criteria or strategies.
On a side note, I wish the name “approval” voting could be changed to “support” voting.
score-stratified-condorcet [10] cardinal-condorcet [9] ranked-condorcet [8] score [7] approval [6] ranked-bucklin [5] star [4] ranked-irv [3] ranked-borda [2] for-against [1] distribute [0] choose-one [0]
-
@cfrank The issue with only ranking candidates you approve is that that gives minority faction voters who don’t like the frontrunners no voice. Unless they’re strategic… but then they basically have to disregard the instructions.
-
@sarawolk yes, the key here is that you rank the candidates that you don’t support as well
this way the support(/approval) aspect is somewhat more extricated from consideration of who the front-runners are—you don’t need to feel compelled to support either one. I think this is very much like an existing system but maybe the runoffs don’t proceed the same way there as I’m describing here.For example, say there are 4 candidates Bush, Nader, Gore, and Vermin Supreme. Maybe Bush and Gore are the suspected front-runners and you want Nader to win. You could submit N|G>B>V (or N|G>V>B, perhaps) and this way you are only supporting Nader, but also indicating your preference among the candidates you would prefer not to win just in case Nader winning isn’t a feasible outcome. Maybe the ballots are like this:
40%: B|N>G>V
30%: N>G|B>V
15%: N|G>B>V
12%: G|N>B>V
3%: V|N>G>BIn this case, we have supports
B: 40%
N: 45%
G: 42%
V: 3%So the top two N and G are put head to head. The majority prefers N to G, so G is eliminated. Next N and B are put head to head, and again the majority prefers N. And so on, so N wins.
On a side note, I wanted to apologize for missing the meeting a while back. Other events converged and I had to prioritize my time but should have announced as much.
-
@cfrank said in Rank with cutoff runoff:
What do you think of this? A ranked ballot is submitted with an approval cutoff. The top two approved candidates are put head-to-head in a majoritarian runoff based on the rankings.
Probably choosing where to put the cutoff is not obvious and becomes intrinsically strategic, and related to the position of the two suspected front-runners in the rank. Maybe you would approve of every candidate you prefer to both suspected front-runners. I think it gets too complicated.
Like STAR, this method would also fail independence of clones. Just clone the most approved candidate and the run-off becomes a non-event.
Chris Benham actually mentioned this method just yesterday.
Say, as I
earlier advocated, the voters rank however many candidates they want to
and give an approval cutoff wherever
they want, and we elect the pairwise winner between the two most
approved candidates. -
@toby-pereira I mentioned this prospect to address independence of clones:
“ Here is a wacky alteration:
The top 2 approved candidates are identified. The runoff between them is not for election, but for elimination, I.e. the front-runner who loses the ranking runoff is removed from the election. The process is repeated over the remaining candidates until a single candidate remains.I think this addresses independence of clones but probably opens a can of worms with other criteria or strategies.”
-
Another STAR alternative, discussed here is to provisionally elect the score winner. Then see if that candidate is "covered" by any others. If so then provisionally elect the highest scoring candidate that covers it. Carry on until the provisionally elected candidate is uncovered.
Basically, candidate A covers candidate B if every candidate that pairwise beats A also pairwise beats B, and at least one candidate pairwise beats B but not A.
-
@toby-pereira that STAR alternative is interesting and seems somewhat similar in spirit to the method I’m suggesting now. I’ll look at it more carefully.
Anyway I think what I’m suggesting also suffers from burial and tactical support.
-
What about This one?
