I think one of the biggest useful changes would be to reform voting so that the public gets more bits of input, by switching to approval or Condorcet style voting.
What do you use currently? Something worse than approval? Tell me it isn’t “First Past the Post”!
Condorcet voting systems seem like a good option. We’ve been using Instant Runoff Voting here (Australia) since before we federated but it seems like Condorcet would be a straightforward upgrade. The principle (‘preference voting’) is the same but Condorcet looks like it would better handle the situation where your first preference is (for example) the 2nd most popular candidate.
What do you use currently? Something worse than approval? Tell me it isn’t “First Past the Post”!
Why are you asking me to lie?
A proportional-representation system just won’t fly in (most of) the U.S. I certainly don’t like the enhanced party-discipline it tends to reinforce.
Although in theory approval is subject to most of the same strategic voting problems as FPTP, STV/IRV, and Borda count, in practice, approval works quite well. It’s simpler to explain and count compared to Condorcet, and for n candidates requires only n counts instead of the n(n-1)/2 counts that Condorcet would.
(I do regularly run votes for my smallish, intelligent gaming group, and there we do use Condorcet to e.g. pick the next game and who’s running it—though usually as nice summary for establishing consensus).
Although in theory approval is subject to most of the same strategic voting problems as FPTP, STV/IRV, and Borda count, in practice, approval works quite well.
You’re comparing approval favorably to IRV along dimensions related to strategic voting? That seems bizarre to me. Thinking of cases in which to vote strategically with IRV is relatively difficult—it very rarely matters and only changes the payoffs marginally. With approval voting strategic voting is more or less necessary to vote effectively. You need to know where to draw the line on what could have otherwise been a preference ordering in order to minimise the loss of your preference information due to the system.
I probably wouldn’t bother with Concorcet if not for the ability to use computers to do the counting. IRV is much simpler to count by hand. “OK guys. This candidate is out. Let’s take this box, cross off the top name and sort them again.”
You’re comparing approval favorably to IRV along dimensions related to strategic voting?
Yep. Strategic voting for IRV becomes relevant as soon as the third-ranked candidate becomes competitive, and essentially gives you first-past-the-post behavior. It’s less likely to encourage strategic voting than FPTP, and this is definitely important in practice, but it still falls under the Gibbard-Satterthwaite theorem. See, for example, http://minguo.info/election_methods/irv/
It’s true that optimally setting a cut-off in approval is part of the strategy. But there is never an incentive to lie and approve a lessor-favored candidate over a more-favored one. The second is far more informationally damaging. (And I think it is sometimes easier to just measure each candidate against a cut-off rather than doing a full ranking.)
I probably wouldn’t bother with Concorcet if not for the ability to use computers to do the counting. IRV is much simpler to count by hand.
I’d describe that slightly differently—Condorcet is easier to count by hand—it’s just the pairwise races that matter. Determining the winner from the counts involves a bit of skull sweat. IRV, the counting proper needs a separate bucket for each permutation, but is easier to analyze by hand and determine the winner. YMMV, on whether this is a useful distinction.
What do you use currently? Something worse than approval? Tell me it isn’t “First Past the Post”!
Condorcet voting systems seem like a good option. We’ve been using Instant Runoff Voting here (Australia) since before we federated but it seems like Condorcet would be a straightforward upgrade. The principle (‘preference voting’) is the same but Condorcet looks like it would better handle the situation where your first preference is (for example) the 2nd most popular candidate.
Why are you asking me to lie?
A proportional-representation system just won’t fly in (most of) the U.S. I certainly don’t like the enhanced party-discipline it tends to reinforce.
Although in theory approval is subject to most of the same strategic voting problems as FPTP, STV/IRV, and Borda count, in practice, approval works quite well. It’s simpler to explain and count compared to Condorcet, and for n candidates requires only n counts instead of the n(n-1)/2 counts that Condorcet would.
(I do regularly run votes for my smallish, intelligent gaming group, and there we do use Condorcet to e.g. pick the next game and who’s running it—though usually as nice summary for establishing consensus).
You’re comparing approval favorably to IRV along dimensions related to strategic voting? That seems bizarre to me. Thinking of cases in which to vote strategically with IRV is relatively difficult—it very rarely matters and only changes the payoffs marginally. With approval voting strategic voting is more or less necessary to vote effectively. You need to know where to draw the line on what could have otherwise been a preference ordering in order to minimise the loss of your preference information due to the system.
I probably wouldn’t bother with Concorcet if not for the ability to use computers to do the counting. IRV is much simpler to count by hand. “OK guys. This candidate is out. Let’s take this box, cross off the top name and sort them again.”
Yep. Strategic voting for IRV becomes relevant as soon as the third-ranked candidate becomes competitive, and essentially gives you first-past-the-post behavior. It’s less likely to encourage strategic voting than FPTP, and this is definitely important in practice, but it still falls under the Gibbard-Satterthwaite theorem. See, for example, http://minguo.info/election_methods/irv/
It’s true that optimally setting a cut-off in approval is part of the strategy. But there is never an incentive to lie and approve a lessor-favored candidate over a more-favored one. The second is far more informationally damaging. (And I think it is sometimes easier to just measure each candidate against a cut-off rather than doing a full ranking.)
I’d describe that slightly differently—Condorcet is easier to count by hand—it’s just the pairwise races that matter. Determining the winner from the counts involves a bit of skull sweat. IRV, the counting proper needs a separate bucket for each permutation, but is easier to analyze by hand and determine the winner. YMMV, on whether this is a useful distinction.