5 ms·

Arrow's theorem[1] demonstrates that no voting system can be "efficient" in the normal sense of the word. That is, no voting system can guarantee: 1. If every

Arrow's theorem[1] demonstrates that no voting system can be "efficient" in the normal sense of the word. That is, no voting system can guarantee: 1. If every voter prefers A over B, the electorate prefers A over B 2. If no voter changes preference between A and B, the electorate's preference of A and B is unchanged 3. No single voter can always change the electorate's preference [1]: https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theorem https://en.wikipedia.org/wiki/Arrow%27s_impossibility_theore...

VodkaHaze 10y ago
That's a bad argument to make, though, because nothing about Arrow's theorem says how much the criteria will be broken, only that they can't be fully satisfied. Moreover, it doesn't apply to "range voting" methods (cardinal information-based voting systems). See here: https://www.reddit.com/r/badeconomics/comments/5a9a1b/arrows_impossibility_theorem_implies_there_is_no/ https://www.reddit.com/r/badeconomics/comments/5a9a1b/arrows...
wutbrodo 10y ago
That's a different issue from what the parent comment is talking about. He's not saying that IRV will be fully efficient, but that the switch itself will be Pareto efficient (which means that each entity is better off, or the same). This actually isn't true, because major party candidates will be worse off, but I think he meant something like Hicks-Kaldor efficient, or positive sum.
- stanleydrew 10y ago
  Yes, I wasn't claiming that there is a perfect voting system. Just that IRV seems strictly better for pretty much all voters than FPTP. I am also seeing several comments that try to take into account the preferences of the candidates themselves, or their parties. Why? Aren't we talking about efficiency from the perspective of the electorate? To the extent that the candidates are themselves members of the electorate, and can vote for themselves, shouldn't we ignore them in this analysis because their preferences are inherently biased?
  - wutbrodo 10y ago
    > I am also seeing several comments that try to take into account the preferences of the candidates themselves, or their parties. Why? Aren't we talking about efficiency from the perspective of the electorate? To the extent that the candidates are themselves members of the electorate, and can vote for themselves, shouldn't we ignore them in this analysis because their preferences are inherently biased? I fully agree with you. I was just being precise in my usage of Pareto efficiency, which includes all agents in a situation. Especially because it's used so often when Hicks-Kaldor improvement is meant (which means that there's the possibility for Pareto improvement after redistribution of gains).
rspeer 10y ago
You missed a criterion: All voters' preferences are expressed through a ranking. As an impractical example, if you could get every voter to tell you exactly how many utils a particular candidate is worth to them (maybe through complex incentives), as a quantity and not as a ranking, you would end up with a perfect voting system. As a more practical example, approval and range voting don't fall victim to Arrow's theorem.
- tzs 10y ago
  > As an impractical example, if you could get every voter to tell you exactly how many utils a particular candidate is worth to them (maybe through complex incentives), as a quantity and not as a ranking, you would end up with a perfect voting system. Would end up with a perfect system, or could end up with a perfect system? I assume that latter, because I can easily design a system that uses that extra data and is far from perfect. Assuming could, not would, my question then is that known? Or is it just known that Arrow's theorem does not apply to such systems, and so we cannot rule out that such a system could be perfect?
tdeck 10y ago
As mentioned in the article, Arrow's theorem only applies to rank-order voting systems (including FPTP and IRV). There are other types of voting systems (e.g. approval voting) that do not require the voter to be strategic in expressing her preference.

VodkaHaze 10y ago

That's a bad argument to make, though, because nothing about Arrow's theorem says how much the criteria will be broken, only that they can't be fully satisfied. Moreover, it doesn't apply to "range voting" methods (cardinal information-based voting systems). See here: https://www.reddit.com/r/badeconomics/comments/5a9a1b/arrows_impossibility_theorem_implies_there_is_no/ https://www.reddit.com/r/badeconomics/comments/5a9a1b/arrows...

wutbrodo 10y ago

That's a different issue from what the parent comment is talking about. He's not saying that IRV will be fully efficient, but that the switch itself will be Pareto efficient (which means that each entity is better off, or the same). This actually isn't true, because major party candidates will be worse off, but I think he meant something like Hicks-Kaldor efficient, or positive sum.

stanleydrew 10y ago
Yes, I wasn't claiming that there is a perfect voting system. Just that IRV seems strictly better for pretty much all voters than FPTP. I am also seeing several comments that try to take into account the preferences of the candidates themselves, or their parties. Why? Aren't we talking about efficiency from the perspective of the electorate? To the extent that the candidates are themselves members of the electorate, and can vote for themselves, shouldn't we ignore them in this analysis because their preferences are inherently biased?
- wutbrodo 10y ago
  > I am also seeing several comments that try to take into account the preferences of the candidates themselves, or their parties. Why? Aren't we talking about efficiency from the perspective of the electorate? To the extent that the candidates are themselves members of the electorate, and can vote for themselves, shouldn't we ignore them in this analysis because their preferences are inherently biased? I fully agree with you. I was just being precise in my usage of Pareto efficiency, which includes all agents in a situation. Especially because it's used so often when Hicks-Kaldor improvement is meant (which means that there's the possibility for Pareto improvement after redistribution of gains).

rspeer 10y ago

You missed a criterion: All voters' preferences are expressed through a ranking. As an impractical example, if you could get every voter to tell you exactly how many utils a particular candidate is worth to them (maybe through complex incentives), as a quantity and not as a ranking, you would end up with a perfect voting system. As a more practical example, approval and range voting don't fall victim to Arrow's theorem.

tzs 10y ago
> As an impractical example, if you could get every voter to tell you exactly how many utils a particular candidate is worth to them (maybe through complex incentives), as a quantity and not as a ranking, you would end up with a perfect voting system. Would end up with a perfect system, or could end up with a perfect system? I assume that latter, because I can easily design a system that uses that extra data and is far from perfect. Assuming could, not would, my question then is that known? Or is it just known that Arrow's theorem does not apply to such systems, and so we cannot rule out that such a system could be perfect?

tdeck 10y ago

As mentioned in the article, Arrow's theorem only applies to rank-order voting systems (including FPTP and IRV). There are other types of voting systems (e.g. approval voting) that do not require the voter to be strategic in expressing her preference.