3 ms·

You never lose by getting more votes - the nice thing about instant run offs is it simple (just rank) and it is close to the ideal outcome. It is not perfect, b

You never lose by getting more votes - the nice thing about instant run offs is it simple (just rank) and it is close to the ideal outcome. It is not perfect, but it is not too bad. In regards stability it is better than proportional voting which tends to splinter the electorate. With instant run offs the candidate that is close to the consensus tends to win.

aninhumer 11y ago
>You never lose by getting more votes Yes you can. Basically the scenario is that by gaining votes from other parties, you can change who is eliminated in the first round, which changes distribution of second round votes. You can argue about how likely this is to happen, but on some level it doesn't matter. If people can't feel absolutely confident that voting for who they actually want won't reduce their vote's impact, many will continue to vote tactically.
- danieltillett 11y ago
  That is not how the system works for in example Australia. What happens is the votes for the lowest ranked candidate are given to all the second choice candidates and the process repeated until you get a candidate with 50% of the vote. There is no way that placing your preferred candidate first can ever result in your vote not being given in the order you choose. If you prefer candidate X above candidate y then candidate y will never get your vote unless X is eliminated.
  - aninhumer 11y ago
    Yes, I understand how IRV works, and indeed I used to believe it was flawless. >There is no way that placing your preferred candidate first can ever result in your vote not being given in the order you choose. No, but it can end up harming your second choice, by eliminating them early, so your vote for them doesn't matter. This is called Favourite Betrayal, and it's a serious problem with IRV: https://www.youtube.com/watch?v=JtKAScORevQ https://www.youtube.com/watch?v=JtKAScORevQ And as a quick worked example, if you have 4 people voting A>B>C, 4 people voting C>B>A and 3 people voting B>A>C, then A wins. If one person voting C>B>A changes their vote to B>C>A, then B wins. So by putting C first, they were harming their second choice B.
    - danieltillett 11y ago
      These type of contrived examples appear to occur rarely in practice. I would love to see some bootstrap simulations on real IRV elections results to see how sensitive IRV is to these pathological outcomes. Most of the time you are only interested in which of the major candidates gets your second choice.
      - aninhumer 11y ago
        I don't know if they're that contrived. The scenario is that you have a closely matched two party system (as most FPTP places do) and a third party which is close in popularity to one of the mainstream ones on first votes, but is not universally supported by the remaining mainstream voters over the other mainstream party. That doesn't just sound plausible, it sounds like exactly the pattern of growth for a third party that you'd expect.
nitrogen 11y ago
I'm drawing my conclusions from the visualizations here: http://zesty.ca/voting/sim/ http://zesty.ca/voting/sim/ (search for "nonmonotonicity"). Maybe I should have said public opinion. Shifting public opinion closer to your position can make you lose in IRV; I'm not sure if there's an actual vote count increase as well. From those visualizations and my reading elsewhere it looks like approval voting is the closest to a Voronoi-style partitioning of the opinion space without being subject to strategic voting.
- edraferi 11y ago
  That's a great article, thanks! I feel like I understand the tradeoffs between voting methods much better now.
- danieltillett 11y ago
  There is no perfect electrol system, just systems of more or less degrees of badness. Any system chosen is always a balance between simplicity and correctness.
  - aninhumer 11y ago
    Honestly, I've yet to be convinced of a serious problem with Approval/Range Voting. They even survive Arrow's Theorem by not being ranked systems. And there's obviously some variance of complexity between them, but even fine grained Range Voting isn't that hard to understand.
    - ClayShentrup 11y ago
      Score Voting and Approval Voting are clearly the best alternative systems for single-winner elections (for multi-winner elections, many people support some form of Proportional Representation). The problem is that FairVote, which advocates IRV (apparently because it's a useful steeping stone to the their REAL goal of STV), is run by a professional liar named Rob Richie, and he disseminates a massive amount of false and misleading information about voting systems. E.g. https://sites.google.com/a/electology.org/www/fact-check https://sites.google.com/a/electology.org/www/fact-check

aninhumer 11y ago

>You never lose by getting more votes Yes you can. Basically the scenario is that by gaining votes from other parties, you can change who is eliminated in the first round, which changes distribution of second round votes. You can argue about how likely this is to happen, but on some level it doesn't matter. If people can't feel absolutely confident that voting for who they actually want won't reduce their vote's impact, many will continue to vote tactically.

danieltillett 11y ago
That is not how the system works for in example Australia. What happens is the votes for the lowest ranked candidate are given to all the second choice candidates and the process repeated until you get a candidate with 50% of the vote. There is no way that placing your preferred candidate first can ever result in your vote not being given in the order you choose. If you prefer candidate X above candidate y then candidate y will never get your vote unless X is eliminated.
- aninhumer 11y ago
  Yes, I understand how IRV works, and indeed I used to believe it was flawless. >There is no way that placing your preferred candidate first can ever result in your vote not being given in the order you choose. No, but it can end up harming your second choice, by eliminating them early, so your vote for them doesn't matter. This is called Favourite Betrayal, and it's a serious problem with IRV: https://www.youtube.com/watch?v=JtKAScORevQ https://www.youtube.com/watch?v=JtKAScORevQ And as a quick worked example, if you have 4 people voting A>B>C, 4 people voting C>B>A and 3 people voting B>A>C, then A wins. If one person voting C>B>A changes their vote to B>C>A, then B wins. So by putting C first, they were harming their second choice B.
  - danieltillett 11y ago
    These type of contrived examples appear to occur rarely in practice. I would love to see some bootstrap simulations on real IRV elections results to see how sensitive IRV is to these pathological outcomes. Most of the time you are only interested in which of the major candidates gets your second choice.
    - aninhumer 11y ago
      I don't know if they're that contrived. The scenario is that you have a closely matched two party system (as most FPTP places do) and a third party which is close in popularity to one of the mainstream ones on first votes, but is not universally supported by the remaining mainstream voters over the other mainstream party. That doesn't just sound plausible, it sounds like exactly the pattern of growth for a third party that you'd expect.

nitrogen 11y ago

I'm drawing my conclusions from the visualizations here: http://zesty.ca/voting/sim/ http://zesty.ca/voting/sim/ (search for "nonmonotonicity"). Maybe I should have said public opinion. Shifting public opinion closer to your position can make you lose in IRV; I'm not sure if there's an actual vote count increase as well. From those visualizations and my reading elsewhere it looks like approval voting is the closest to a Voronoi-style partitioning of the opinion space without being subject to strategic voting.

edraferi 11y ago
That's a great article, thanks! I feel like I understand the tradeoffs between voting methods much better now.
danieltillett 11y ago
There is no perfect electrol system, just systems of more or less degrees of badness. Any system chosen is always a balance between simplicity and correctness.
- aninhumer 11y ago
  Honestly, I've yet to be convinced of a serious problem with Approval/Range Voting. They even survive Arrow's Theorem by not being ranked systems. And there's obviously some variance of complexity between them, but even fine grained Range Voting isn't that hard to understand.
  - ClayShentrup 11y ago
    Score Voting and Approval Voting are clearly the best alternative systems for single-winner elections (for multi-winner elections, many people support some form of Proportional Representation). The problem is that FairVote, which advocates IRV (apparently because it's a useful steeping stone to the their REAL goal of STV), is run by a professional liar named Rob Richie, and he disseminates a massive amount of false and misleading information about voting systems. E.g. https://sites.google.com/a/electology.org/www/fact-check https://sites.google.com/a/electology.org/www/fact-check