Getting the Percieved Advantage of IRV but with Equal Votes and Precinct Summability

One of the characteristics of IRV is that a round cannot be tallied until after the prior round has been tallied. Maybe there’s a way around that while still somewhat addressing the IRV advocates’ criticism of Score that any support you give to your compromise candidate hurts the chances of your favorite winning.

Suppose we had two rounds of tallying, and the ballot would consist of a Score ballot for each of the two rounds. The first round would eliminate half the candidates. So for that round, the voters could concentrate on trying to eliminate their most hated candidates. They might well vote in Approval style, and maybe an Approval ballot is all that is needed for the first round. An elaboration would use more than two rounds, but the count of rounds would still be fixed from the start, and the counts of eliminations for each round adjusted so that at least two candidates would survive to the final round. Maybe it would go logarithmically, with each round eliminating about half the remaining candidates. Or linearly, with each round eliminating about the same number of candidates, with the possible exception of the last round. Anyway, the precincts could publish the results round by round; there would be no need to wait for a central decision between starting to tally rounds. The calculation of the final result would ignore for each round, the scores of candidates already eliminated in a prior round.

This would turn into a festival where the major parties try to run a majority of candidates which all get approved and the second round is all Republicans.

Why not just use STAR? Precincts can publish the pairwise table which requires only O(C^2) information (C=num candidates) and does not leak individual votes.


Even though STAR has been published, IRV advocates continue to advocate for IRV.

Let me modify this proposal – the rounds other than the final would use RRV to promote candidates.

I’m confused if you are saying people have to go back to the polls and vote again. If so, that’s a tough sell.

Even just filling out two ballots differently is a tough sell. That sounds like an awful user experience.

If you want something to sell to the IRV people, I’d be thrilled with a ranked method that chose the Condorcet candidate if one existed, and if none exists, chose the member of the Smith set that wins an IRV tally. FairVote’s site even claims that they consider the Condorcet criterion important. They say: (*)

We also consider the Condorcet criterion to be important. This is the property that the candidate that would win a head-to-head race against every other candidate should always win. While (IRV), approval, and score voting may fail to elect the Condorcet candidate, in practice (IRV) has done so in virtually every single election.

Interestingly, they go on to say:

Due to strategic vulnerabilities of Condorcet methods, including later-no-harm, and the additional complexity Condorcet requires to resolve cycles, we strongly prefer (IRV) for political elections.

So… use your IRV formula for resolving cycles, and you can’t complain about the complexity of resolving cycles. And you don’t have to apologize for the fact that sometimes it won’t elect the Condorcet candidate.

What am I missing?

You might be interested in this:

What you described sounds like Woodall’s method in that link. Arguably, Benham’s method is simpler, closer to IRV, and gives just as justifiable a result as Woodall.

Thanks, yeah, I don’t have a strong preference. They both pick the Condorcet winner if s/he exists, if not falls back to IRV. Woodall seems simpler to me only because it can be explained in less words (I’d think). Since I would estimate that 99% of real-world elections with more than a few hundred voters will have a Condorcet winner, I think it makes sense to describe it as “Condorcet with IRV fallback” rather than “IRV with Condorcet attempted between rounds,” but it doesn’t really matter.

A way to generalize from your warning about a festival, even though you mention certain known parties, and even though you bring the warning in the context of my suggesting a system with two rounds of tallying and no return to the polls, is to read into it that for any multiround approach, irrespective of whether the voters return to the polls for later rounds or whether the rounds apply in tallying one polling (or some combination), but where early rounds can eliminate candidates, the risk exists of a faction gaining unfair advantage by working to promote into a later round, one or more candidates for the purpose of squeezing out of those later rounds, a candidate who could win under straight Range but whom they don’t like. Having been alerted to this risk, I think it might be real and I tentatively think the only way to avoid it is to avoid systems that eliminate candidates in early rounds. You agree?