Interesting argument for Ranking over Cardinal methods

I heard an interesting argument for Ordinal methods over Cardinal methods.

If you get people to rank and score then you show them the other voters ballots then people are less likely to want to change their ranking. This is not a great argument but I can see how it would be persuasive. Does anybody know a good counter for the layperson? I think pointing out that this is not realistic is not enough.

Is this because of voters learning to strategize, or something else?
If it’s the former, then any amount of polling would do the same for them. We already know that Score and Approval turn into honest Condorcet methods under total strategy, and it isn’t hard to imagine voters being somewhat strategic and boosting candidates they prefer over the frontrunners in a bid to… well, beat those frontrunners. So the effect is pretty much guaranteed unless Score voters choose to stay honest for whatever reasons, meaning that there’s no real concern over this, since everyone will change their strategy before the election, or at least, after the first elections.
Edit: To correct myself, it’s actually the case that you have to give the maximal score/approve the Condorcet winner if you think they’re better than the expected value of the winner. As best as I can tell, this would likely mean the average of the Top 2’s utilities, which is a bit more complicated, and makes electing a Condorcet winner in Approval or Score less of a sureshot, and makes Score’s added expressivity valuable, though I see more room for a Condorcet winner in STAR, since its runoff is primed for one.
This is why Approval and STAR really take the cake over Score though, if you follow this one argument. Approval forces everyone’s votes to be equal, while STAR’s runoff makes it fairly difficult to be “strategic” in the face of new polling/voter strategy information.
Also, my counter to people supporting ranking is this: if 10% of the voters rank a candidate 1st place, and 80% rank that candidate 2nd place, the results will only report that candidate got 10% of the vote, crushing candidates who are good compromises/popular. Scoring and approving are the only way to show how much support each candidate got, independently, and STAR basically falls in this category, since all candidates other than the Top 2 are measured by their scored support only.

1 Like

It’s not true, since no reasonable ranked choice system always incentivizes honest rankings.

1 Like

It is basically that if you know all other voter’s votes then you can vote optimally strategically. In most cases this means to bullet vote for a single candidate where their honest utility is not that. Since IRV is hard to compute mentally most people will not be able to sort out this strategy in that system.

example:
Initial vote A:5 B:4 C:0

If then B wins by 2 points over A then the voter would want to change their vote to A:5 B:0 C:0 so that their favourite would win. The reason they initially voted non zero for B was to guard against C. If they know this is not a threat then they can safely bullet vote. This would exist in a ranking system where they vote A>B>C and B wins. Even if C is eliminated first changing the ranking is does not ensure A to win.

I agree that polling would do the same to some extent. The point is that this is fake situation which FairVote can use to trick people. I know that honest ranking is not really expected in the first place but in a toy example they give like ranking food they will likely be pretty honest.

1 Like

That is only true of IRV. Other ranking methods do take your entire ballot into account.

But polling only gives a general idea. You would not be able to know the exact outcome down to a ballot’s worth of points.

Perhaps a point can be made here clarifying that in the best-case scenario, Score elects a utilitarian winner, while in a worst-case scenario where everyone tries to strategize, Score tends towards Condorcet winners, which is the best that RCV can likely do? After all, the only way A voters are really “safe” here is if they know for sure B-top voters won’t end up giving A a 0 to preempt this sort of strategy from A-top voters. And B-top voters themselves will have to know for sure that A-top voters won’t give B a 0 anyways, meaning both A and B get sunk. So in the end, this sort of strategy still has dangers. This is probably the real reason why it isn’t necessary to replace STAR Score with Score STAR, by the way; one-sided strategy is likely to get caught or even backfire quickly.
STAR advocates themselves made a similar argument at https://www.equal.vote/fv:

Fundamentally, for the “burying” tactic FairVote describes to work, voters from opposing factions have to gang up on a well-liked consensus candidate, by supporting the opponent they really don’t like higher on the ballot than their true second choice. Yet in reality, if either of those factions actually believe that the opposing faction is going to adopt this strategy, their own best play is to simply vote honestly, in order to give their own favorite the best chance of winning against the (now diminished) consensus choice.

This is the deepest flaw in FairVote’s hypothesis: the proposed tactic requires voters of true opponents to work together to be dishonest on their ballots, yet if one faction decides to be honest instead, the honest faction will gain the significant upper hand . For this reason alone, we see this hypothetical tactic as an obvious non-starter.

But the actual results will only likely mention a) the Condorcet winner’s pairwise victories over the top candidates and b) the first-preference votes of all candidates. So unless we are lucky enough to have a 3rd party be a Condorcet winner right off the bat, there will still be a strong element of “that candidate doesn’t have enough votes to ‘win’.”

2 Likes

STAR and IRV fail in this scenario:
51% A5 B4 C0
49% C5 B3 A0

STAR does not fail in this scenario BUT it shows that STAR is not clone-immune (but the effect is limited, as opposed to e.g. Borda):
51% A5 B4 B’3 C0
49% C5 B’4 B3 A0

IRV also fails this scenario (but STAR does not):
39% A5 > B3 > C0
10% B5 > A1 > C0
10% B5 > C1 > A0
41% C5 > B4 > A0

I’ll summarize my response to this in the other thread:
In the first example, the minority has somewhat of an incentive to run two consensus candidates to crowd out the runoff, which would make STAR work.
In the second example, IRV would also make it harder for people to see the support the consensus candidate actually has, since it only shows 1st preferences in results, meaning B would appear to have 20% support, rather than the more accurate 76.2% of the points that STAR would reveal. That not only means people would write off the consensus candidate in IRV, they might not even rank them.

Edited.

Yeah, and I also consider that third scenario more realistic than the first two. In small elections like these, the centrists will probably have a decently strong amount of “core support” (20% in this case). In bigger elections that may not happen, although even then it would seem like the best consensus candidates will probably be those who do have core support anyway.

I think that for someone to truly be a utilitarian-style compromise, they couldn’t possibly be a lot of people’s 1st preference. Maybe they’d be close in utility to 20% of the population’s 1st preferences though, making them a close 2nd.

Yes, but there will be many voters who genuinely find moderates better than the two factions and so the moderate is their first choice. The US is 25% liberal and 36% conservative according to a Gallup poll. Those do not add up to anything close to 100%.

I suppose it is a matter of how many candidates run, how many shades of “moderate” there are on offer, and whether there really would be two factions at all in cardinal systems, or if it would be closer to majority rule on separate, individual issues. I still hesitate to call it “core support” from that “moderate majority”, because a lot of those “moderates” are really folks who are liberal on some issues and conservative on others, so even if they do prefer any compromise candidate over either faction, they may yet have much stronger preference for their favorite. It’s more like a “I support you strongly because you have the best chance of beating the candidate(s) I don’t like”, which we already strongly have with today’s duopoly.

Personally, I think that the strategic voting of changing a 10,6,0 vote into 10,10,0 feels less damaging to an election than if you changed it into 5,0,10 (which you have no reason to do in Score, but you may do something similar in ranking elections).

1 Like

The problem is, most people would never think to dishonestly rank, while most will see the min/max scoring as obvious, which somewhat hurts the legitimacy/“fairness” of scoring, even if ranking is inferior in quality. Another way STAR has a propaganda advantage.

Australia. (Is two-party dominated yet third parties win seats in non-IRV elections; hence a good number of voters rank dishonestly.)

But I feel like that simply degenerates it into Approval, which is an OK system, and only for those people who actually choose to min-max.

IIRC, the percentage of 1st-preference votes in IRV elections match the percentage of seats they win in PR closely, meaning IRV just doesn’t favor “low core support” and eliminates them. On top of that, we have never heard of a major, public Condorcet failure in Australia that might provoke strategizing. It is possible some voters got tired of voting 3rd party knowing they’ll lose, so they just don’t mark them on their ballot (call it “top-truncation”?)

Others might say it’s an unfairly earned strategic advantage, even if it isn’t, or that the system favors some voters over others (which is the whole point of utilitarianism, but not everyone buys into that interpretation of politics.) Anyways, that’s where STAR’s “I’ll turn your ballot which gave the runner-up a 3 and the top candidate a 4 into a 0 for the runner-up and a 5 for the other candidate!” method is seen as “democracy!” rather than “why is the method misinterpreting my ballot!” People haven’t fully bought into utilitarian thinking yet, and it may be too much to expect that for now.