4 ms·

To expand on your point: there is one specific point that CGP Grey gets wrong in his IRV/RCV videos, which is that he says the system is not worse than first-pa

by elihu 6y ago

To expand on your point: there is one specific point that CGP Grey gets wrong in his IRV/RCV videos, which is that he says the system is not worse than first-past-the-post in any way. On the contrary, FPTP adheres to the monotonicity criterion, which means you can't harm your preferred candidate's chances of winning by voting for them. In RCV, you actually can cause your preferred candidate to lose by ranking them too high in some situations. As far as I know, approval voting is strictly equal to or better than FPTP on any reasonable voting system criterion, which I think makes it a better option. Also in its favor is that it's simpler and doesn't tend to eliminate compromise candidates who weren't anyone's first choice. Range and STAR are also both good systems.

godelski 6y ago
What criteria is Approval not better than Plurality? Only one I can think of is simplicity (although I'd argue that plurality is only slightly simpler). But approval doesn't have the favorite betrayer like ordinal (RCV/IRV/Condorcet) does. As far as I'm aware, cardinal systems are all like this. They are much more resistant to spoilers.
- elihu 6y ago
  I think you're in agreement with me, unless I somehow said something backwards. I think that approval is equal to or better than plurality (FPTP) in every criteria I'm aware of. (One criteria you could argue about is later-no-harm; IRV passes, but approval voting does not. It's kind of silly though to say that it's safe in RCV to put lower ranked options below your first choice when in RCV it isn't even always safe to put your first choice first. FPTP adheres to later no harm by not allowing you to vote for anyone but your first choice in the first place, so one could reasonably say that Approval isn't really worse than FPTP, except as a sort of technicality.)
  - godelski 6y ago
    I believe we're in agreement. I was coming to your defense since there seemed to be some confusion among readers who didn't like your extension. Maybe wording? But yeah, I think on basically every metric cardinal is better than ordinal methods which are better than plurality. The whole IRV push seems odd to me when approval (or STAR) is both simpler, responds better to spoilers (the main concern?), and obtains a substantially higher VSE. I don't think IRV solves enough of the problems to make a meaningful difference (I've brought up Australia a lot in this thread).
- dane-pgp 6y ago
  > I'd argue that plurality is only slightly simpler I've actually heard it argued that Approval is simpler (for the voter) because all valid FPTP ballots are valid in Approval, but not the converse. There is definitely some benefit to a system that makes invalid ballots less likely (although that doesn't guarantee that the voter's intention is captured). The difference in complexity for counting ballots, however, should not be understated. For elections with a paper trail, if you are putting each ballot in a single pile with other similar ballots, then Approval voting can require hundreds of piles. This can delay the process and add to the cost. Jurisdictions which rely on electronic vote counting can avoid this, but that introduces its own trust/legitimacy problems.
  - godelski 6y ago
    > I've actually heard it argued that Approval is simpler (for the voter) because all valid FPTP ballots are valid in Approval, but not the converse. Voting systems can actually be optimal in two candidate systems. It is when you introduce more candidates that things get complex. > For elections with a paper trail, if you are putting each ballot in a single pile with other similar ballots, then Approval voting can require hundreds of piles. This would be poor counting. You need the same number of ballots as FPTP. You just sum multiple columns instead of a single column. The process isn't much different. This is actually the major advantage of approval over even other simple but better systems like score or STAR. It is dead simple to count and understand. Compare any of these to IRV/RCV (what the post is about) and they've got huge complexity problems. But Ireland is able to handle the many rounds of counting you need for IRV and they do it with paper ballots. Remember that Ireland and Australia already have IRV and I'm pretty sure Australia didn't have computers a hundred years ago. (I don't advocate for IRV btw)
    - dane-pgp 6y ago
      > This would be poor counting. You need the same number of ballots as FPTP. You just sum multiple columns instead of a single column. I'm not saying you need more ballots to implement Approval voting, I'm saying that FPTP has the nice property (when counting paper ballots) that the votes can be put into piles, where the weight of each pile is roughly proportional to the number of votes for the candidate whose votes are in that pile. Of course, FPTP votes aren't actually counted using measuring scales, but the fact that this isn't possible with Approval voting is an indication that "just sum multiple columns" is a slightly more laborious and error-prone step than you make it sound, at least for humans counting paper ballots (which I believe is a necessary property of trustworthy elections).

by elihu 6y ago

godelski 6y ago

What criteria is Approval not better than Plurality? Only one I can think of is simplicity (although I'd argue that plurality is only slightly simpler). But approval doesn't have the favorite betrayer like ordinal (RCV/IRV/Condorcet) does. As far as I'm aware, cardinal systems are all like this. They are much more resistant to spoilers.

elihu 6y ago
I think you're in agreement with me, unless I somehow said something backwards. I think that approval is equal to or better than plurality (FPTP) in every criteria I'm aware of. (One criteria you could argue about is later-no-harm; IRV passes, but approval voting does not. It's kind of silly though to say that it's safe in RCV to put lower ranked options below your first choice when in RCV it isn't even always safe to put your first choice first. FPTP adheres to later no harm by not allowing you to vote for anyone but your first choice in the first place, so one could reasonably say that Approval isn't really worse than FPTP, except as a sort of technicality.)
- godelski 6y ago
  I believe we're in agreement. I was coming to your defense since there seemed to be some confusion among readers who didn't like your extension. Maybe wording? But yeah, I think on basically every metric cardinal is better than ordinal methods which are better than plurality. The whole IRV push seems odd to me when approval (or STAR) is both simpler, responds better to spoilers (the main concern?), and obtains a substantially higher VSE. I don't think IRV solves enough of the problems to make a meaningful difference (I've brought up Australia a lot in this thread).
dane-pgp 6y ago
> I'd argue that plurality is only slightly simpler I've actually heard it argued that Approval is simpler (for the voter) because all valid FPTP ballots are valid in Approval, but not the converse. There is definitely some benefit to a system that makes invalid ballots less likely (although that doesn't guarantee that the voter's intention is captured). The difference in complexity for counting ballots, however, should not be understated. For elections with a paper trail, if you are putting each ballot in a single pile with other similar ballots, then Approval voting can require hundreds of piles. This can delay the process and add to the cost. Jurisdictions which rely on electronic vote counting can avoid this, but that introduces its own trust/legitimacy problems.
- godelski 6y ago
  > I've actually heard it argued that Approval is simpler (for the voter) because all valid FPTP ballots are valid in Approval, but not the converse. Voting systems can actually be optimal in two candidate systems. It is when you introduce more candidates that things get complex. > For elections with a paper trail, if you are putting each ballot in a single pile with other similar ballots, then Approval voting can require hundreds of piles. This would be poor counting. You need the same number of ballots as FPTP. You just sum multiple columns instead of a single column. The process isn't much different. This is actually the major advantage of approval over even other simple but better systems like score or STAR. It is dead simple to count and understand. Compare any of these to IRV/RCV (what the post is about) and they've got huge complexity problems. But Ireland is able to handle the many rounds of counting you need for IRV and they do it with paper ballots. Remember that Ireland and Australia already have IRV and I'm pretty sure Australia didn't have computers a hundred years ago. (I don't advocate for IRV btw)
  - dane-pgp 6y ago
    > This would be poor counting. You need the same number of ballots as FPTP. You just sum multiple columns instead of a single column. I'm not saying you need more ballots to implement Approval voting, I'm saying that FPTP has the nice property (when counting paper ballots) that the votes can be put into piles, where the weight of each pile is roughly proportional to the number of votes for the candidate whose votes are in that pile. Of course, FPTP votes aren't actually counted using measuring scales, but the fact that this isn't possible with Approval voting is an indication that "just sum multiple columns" is a slightly more laborious and error-prone step than you make it sound, at least for humans counting paper ballots (which I believe is a necessary property of trustworthy elections).