No, Arrow’s impossibility theorem says that all voting systems have flaws, if you agree with the things it defines as flaws.
The most doubtful of the flaws is the “Independence of irrelevant alternatives” criterion, which says that: if you run two elections in which voters change rank candidates A and B the same with respect to one another, the elections will both rank those candidates the same.
The problem with this is that if voters change their ranking for some other candidate C it can end up affecting the outcome for candidates A and B, when arguably it shouldn’t. But this makes less sense if you realise that aggregating voter preferences can end up implying that candidate A is better than B who is better than C… who is better than A. This setup makes it impossible to maintain the principle.
Arrow’s Theorem only really applies to Ordinal voting systems like RCV and Plurality.
Cardinal systems like STAR, Score, and Approval are all immune, but have other quirks, but seem to better represent the desires of an aggregate population.
I think people are wary of systems where you assign a relative score for good reason. Is my favourite party twice as good as the major party that kind represents my views, or is it only a little bit better? It’s kind of impossible to make those judgements well IMO. In STAR in particular, what you actually want to do is rank your preferences, but if there are many candidates are forced not to because you only have five scores available.
Also you seem to have replied about STAR to just about every comment in the thread… maybe chill? lol…
Mate, I know plenty about voting systems. You don’t need to rattle them off.
As to “forcing you to rank candidates” that’s hogwash.
In STAR, if you have four candidates and you feel like two are pretty similar, you might give them both a three. But if they then make it to the runoff, you’ve no longer expressed a preference between them. Do you prioritise expressing your preferences in the first round (where you might not want to help either of them win against your favourite whom you awarded 5) or in the runoff?
Forcing people to use a bad system because you think they aren’t smart enough to rate someone on a scale of 0-5 is kind of mind-boggling.
Who is “forcing” people to use a “bad” system? Which bad system? Do you think that, because I don’t like STAR which you’ve spammed the entire thread with, I must be in love with FPTP? Let’s take a deep breath here…
Sounds like FUD for any voting system. Security comes from other aspects. Lots of elections use Instant Runoff Voting and STV which has similar properties, without security problems.
spoiler effect, monotonicity issues
Listing these issues is a bit pointless given impossibility theorems. No electoral system can be both Condorcet compatible (always elects the candidate who beats all other candidates in a head-to-head, if such a candidate exists) without also potentially rewarding abstentions (i.e. there are situations where changing from not voting to entering a vote which prefers candidate A to candidate B can cause candidate B to win instead of A). STAR voting satisfies neither principle. IRV satisfies the property that if you switch from ranking A > B > C to A > C > B this can never cause A to lose - STAR voting does not (this is what I outlined above, or part of it).
All voting systems are compromises. But this last issue gets to the real heart of it: STAR voting has this issue precisely because it, as a scoring system, is explicitly saying “one hundred people each assigning this candidate a 1 indicates they should win over a candidate who got 49 2s and no other votes”. Ranking methods are denying the ability to trade off many weak preferences against fewer strong preferences and go only by rankings. Each leads to different issues; you need to go more into those rather than just focus on a list of mathematical properties.
IRV is significantly better than FPTP because the spoiler effect is much less, possibly below the threshold where most voters would actually attempt to vote tactically. STV is significantly better than IRV because it is much more proportional. MMP is even better because it’s simpler and arguably has better local ties between representatives and electorates. These are the real issues - mathematical properties are interesting but not the final word.
Okay, first, the security issues of RCV are well known, but often denied. The votes must be counted in a centralized location. They cannot be tallied at the polling locations. This means you must transport the ballots. RCV is the only voting system that requires the ballots to be transported (or scanned and transmitted) This introduces numerous security issues that do not exist in other voting systems, not even in plurality.
Take the 2021 NYC mayoral election, where there were 100K extra votes. This was due to a screwup in mixing test ballots with real ballots, which was a screw-up in how the test ballots were made. Nevertheless, the way it was caught was the winning candidate looked at the polling numbers and spotted a discrepancy. He was still the winner afterward.
An actual malicious actor could easily steal an election, and no one would know it happened.
That alone is terrifying.
As to Arrow’s Impossibility Theorem, that only applies to Ranked, i.e. Ordinal, voting systems. STAR is a Cardinal voting system and is largely immune to Arrow’s Theorem.
And again, the lie about the spoiler effect. RCV still has the spoiler effect, and it’s even worse for enforcing a two party dominance than FPtP.
The main thing to understand is that RCV/IRV whatever you call it, is still a series of FPtP elections on a single ballot.
You cannot fix the problems of plurality by iterating plurality.
Every general election in the UK involves transporting ballot papers to a central location (per constituency). This isn’t an issue.
You don’t actually have to count IRV ballots in a central place; it just requires coordination for the multiple rounds: you can count the first-choice votes in many locations, sum them, determine who is eliminated then, still in those separate locations, count again, now transferring first choice votes for the eliminated candidate, and repeat.
STV is used all over the world and requires the same processes as IRV.
As to Arrow’s Impossibility Theorem
I didn’t mention Arrow’s theorem in the comment you replied to. The compromise I mentioned applies to scoring systems just as much as ranking systems.
And again, the lie about the spoiler effect. RCV still has the spoiler effect, and it’s even worse for enforcing a two party dominance than FPtP.
Citation needed.
The main thing to understand is that RCV/IRV whatever you call it, is still a series of FPtP elections on a single ballot.
You cannot fix the problems of plurality by iterating plurality.
But… you can. Some of them, anyway. If there are three parties running, A, a similar party B, and Z, where you prefer A > B > Z, then in a FPTP election you have to choose whether to vote for A or B, and if A is the third-most-popular, it’s a bad idea to. With IRV you can rank them A > B > Z, the third-most-popular party A probably gets eliminated, but you still express your opinion B > Z. That situation is, in FPTP, incredibly common.
The situations in which the introduction of option B negatively affects option A in IRV are substantially less common than in FPTP because of this: B has to be significantly more popular, to the point where enough people switch their first choice votes from A to B, resulting in A being eliminated, but B must not be so popular that it couldn’t win head-to-head against Z.
Calling IRV “a series FPTP elections” ignores the important condition under which those successive counts happen.
No, Arrow’s impossibility theorem says that all voting systems have flaws, if you agree with the things it defines as flaws.
The most doubtful of the flaws is the “Independence of irrelevant alternatives” criterion, which says that: if you run two elections in which voters change rank candidates A and B the same with respect to one another, the elections will both rank those candidates the same.
The problem with this is that if voters change their ranking for some other candidate C it can end up affecting the outcome for candidates A and B, when arguably it shouldn’t. But this makes less sense if you realise that aggregating voter preferences can end up implying that candidate A is better than B who is better than C… who is better than A. This setup makes it impossible to maintain the principle.
Arrow’s Theorem only really applies to Ordinal voting systems like RCV and Plurality.
Cardinal systems like STAR, Score, and Approval are all immune, but have other quirks, but seem to better represent the desires of an aggregate population.
I think people are wary of systems where you assign a relative score for good reason. Is my favourite party twice as good as the major party that kind represents my views, or is it only a little bit better? It’s kind of impossible to make those judgements well IMO. In STAR in particular, what you actually want to do is rank your preferences, but if there are many candidates are forced not to because you only have five scores available.
Also you seem to have replied about STAR to just about every comment in the thread… maybe chill? lol…
STAR is great, but I can also talk about Approval. That one is dead simple.
The ballot is the same as a FPtP ballot except instead of it saying “mark one” it says something like “mark one or more”.
Approval says vote for as many people as you want, and if any of them win, you’ll be happy. Or not. I’m not the boss of you.
As to “forcing you to rank candidates” that’s hogwash.
Forcing people to use a bad system because you think they aren’t smart enough to rate someone on a scale of 0-5 is kind of mind-boggling.
Mate, I know plenty about voting systems. You don’t need to rattle them off.
In STAR, if you have four candidates and you feel like two are pretty similar, you might give them both a three. But if they then make it to the runoff, you’ve no longer expressed a preference between them. Do you prioritise expressing your preferences in the first round (where you might not want to help either of them win against your favourite whom you awarded 5) or in the runoff?
Who is “forcing” people to use a “bad” system? Which bad system? Do you think that, because I don’t like STAR which you’ve spammed the entire thread with, I must be in love with FPTP? Let’s take a deep breath here…
RCV is literally a bad system. See my other comments about it.
But it has things like the spoiler effect, monotonicity issues, security issues, and more.
It’s a bad system, that’s already failed multiple times in real world election, and people are still pushing for it to have wider implementation.
Sounds like FUD for any voting system. Security comes from other aspects. Lots of elections use Instant Runoff Voting and STV which has similar properties, without security problems.
Listing these issues is a bit pointless given impossibility theorems. No electoral system can be both Condorcet compatible (always elects the candidate who beats all other candidates in a head-to-head, if such a candidate exists) without also potentially rewarding abstentions (i.e. there are situations where changing from not voting to entering a vote which prefers candidate A to candidate B can cause candidate B to win instead of A). STAR voting satisfies neither principle. IRV satisfies the property that if you switch from ranking A > B > C to A > C > B this can never cause A to lose - STAR voting does not (this is what I outlined above, or part of it).
All voting systems are compromises. But this last issue gets to the real heart of it: STAR voting has this issue precisely because it, as a scoring system, is explicitly saying “one hundred people each assigning this candidate a 1 indicates they should win over a candidate who got 49 2s and no other votes”. Ranking methods are denying the ability to trade off many weak preferences against fewer strong preferences and go only by rankings. Each leads to different issues; you need to go more into those rather than just focus on a list of mathematical properties.
IRV is significantly better than FPTP because the spoiler effect is much less, possibly below the threshold where most voters would actually attempt to vote tactically. STV is significantly better than IRV because it is much more proportional. MMP is even better because it’s simpler and arguably has better local ties between representatives and electorates. These are the real issues - mathematical properties are interesting but not the final word.
Okay, first, the security issues of RCV are well known, but often denied. The votes must be counted in a centralized location. They cannot be tallied at the polling locations. This means you must transport the ballots. RCV is the only voting system that requires the ballots to be transported (or scanned and transmitted) This introduces numerous security issues that do not exist in other voting systems, not even in plurality.
Take the 2021 NYC mayoral election, where there were 100K extra votes. This was due to a screwup in mixing test ballots with real ballots, which was a screw-up in how the test ballots were made. Nevertheless, the way it was caught was the winning candidate looked at the polling numbers and spotted a discrepancy. He was still the winner afterward.
An actual malicious actor could easily steal an election, and no one would know it happened.
That alone is terrifying.
As to Arrow’s Impossibility Theorem, that only applies to Ranked, i.e. Ordinal, voting systems. STAR is a Cardinal voting system and is largely immune to Arrow’s Theorem.
And again, the lie about the spoiler effect. RCV still has the spoiler effect, and it’s even worse for enforcing a two party dominance than FPtP.
The main thing to understand is that RCV/IRV whatever you call it, is still a series of FPtP elections on a single ballot.
You cannot fix the problems of plurality by iterating plurality.
Every general election in the UK involves transporting ballot papers to a central location (per constituency). This isn’t an issue.
You don’t actually have to count IRV ballots in a central place; it just requires coordination for the multiple rounds: you can count the first-choice votes in many locations, sum them, determine who is eliminated then, still in those separate locations, count again, now transferring first choice votes for the eliminated candidate, and repeat.
STV is used all over the world and requires the same processes as IRV.
I didn’t mention Arrow’s theorem in the comment you replied to. The compromise I mentioned applies to scoring systems just as much as ranking systems.
Citation needed.
But… you can. Some of them, anyway. If there are three parties running, A, a similar party B, and Z, where you prefer A > B > Z, then in a FPTP election you have to choose whether to vote for A or B, and if A is the third-most-popular, it’s a bad idea to. With IRV you can rank them A > B > Z, the third-most-popular party A probably gets eliminated, but you still express your opinion B > Z. That situation is, in FPTP, incredibly common.
The situations in which the introduction of option B negatively affects option A in IRV are substantially less common than in FPTP because of this: B has to be significantly more popular, to the point where enough people switch their first choice votes from A to B, resulting in A being eliminated, but B must not be so popular that it couldn’t win head-to-head against Z.
Calling IRV “a series FPTP elections” ignores the important condition under which those successive counts happen.