Minneapolis' instant-runoff voting plan may be unconstitutional?
A funny article today in the Star Tribune:
http://www.startribune.com/587/story/1433875.html Minneapolis voters have approved a plan for instant-runoff voting, but the state constitution may not allow it, according to a new attorney general's opinion. The opinion, obtained by the Star Tribune, doesn't explicitly say the Minneapolis system of ranking candidates in order of preference is unconstitutional.
But it concludes that if the closest case to a precedent is followed, instant-runoff voting probably isn't permitted by the state constitution. ... Opponents have threatened a legal challenge, citing a 1915 Minnesota Supreme Court case that struck down a Duluth system that also ranked voters.
The 1915 ruling was against a different rank ballot system called Bucklin which allowed second choice votes to count along with first choice votes, but IRV doesn't allow this. It only counts second choice votes if the first choice is eliminated (just like any runoff process).
If the courts must be involved, I suppose its good to be early, and perhaps a ruling in favor of IRV will reduce the basis for any future rejections of IRV.
The interesting thing for me is, IF something about IRV is found unconstitutional, AND IRV is logically equivalent to a runoff (or top-two primary), then the same unconstitutionality must apply to runoffs or primaries which take candidates off the ballot.
I don't see a constitutional issue with IRV, but there are complaints I take a little more seriously. Specifically, the theoretical possibility that in IRV it is possible to have your candidate LOSE by voting FOR him, and WIN by voting against him. This illogic in election system theory is called nonmonotonicity.
It is possible for the same reason that ANY runoff method allows possible strategic voting. Basically consider a "3-way race". If your favorite is on top, perhaps with 40% support, but needs 50% to win. Then you have two strong rivals, say both around 30% support (being close makes the manipulation possible). Perhaps one rival is "more centrist" and the other is "more extreme" on the opposite political spetrum. You might think your candidate can beat the extreme candidate in the final round, BUT more uncertain against the centrist who will pick up more votes when the extreme candidate is eliminated. SO if a fraction of supporters of your candidate strategically move your votes to the opposite extreme, you can perhaps help the centrist get eliminated.
Interestingly a REAL runoff makes this manipulation easier, and it works in a top-two primary similarly. I can vote in the primary for the opposite extreme, get him into the general election, and then vote my TRUE preference in the general election to beat the weaker competitor.
On the other hand, IRV makes this manipulation more difficult, since you have a rank ballot representing your choices, and you can NOT CHANGE your vote after your stronger rival is beaten. This strategy can still work in IRV more weakly, and only if you trust enough of the centrists supporters will support your favorite second, but the strategy can backfire obviously.
Overall such manipulations are difficult, and require careful analysis of how you expect other voters to act, but I've seen such efforts made, in "party endorsement elections" where delegates vote 10 rounds or more trying to get 60% behind one candidates. Some delegates like to "play" and move their vote around, faking and returning, trying to pretend there's momentum in a direction you like (helping an extremist beat your stronger competitor for instance.) The games happen, whether they actually change anything who knows?
Runoff processes are pretty boring places to play games, and they are mostly a game of dominoes - knowing pretty much the choices will fall in order from the first vote, votes transfering up to stronger candidates in a pretty even and expected way.
The problem I have with IRV is the same problem I have with any runoff process. The hardest to defend aspect of IRV is the "bottom up" elimination, sort of like the endorsement elections, but its just rather arbitrary.
For instance, I've held IRV elections at work for our Halloween constume contests. It is instructive in the difficulties of ties, even if less likely in a larger election. The first round votes were something like this: 5,4,2,1,1,1,1,1,1. There were 17 candidates and 17 voters, so a tough election. Eight voters got zero first rank choices at all, so they are easy to eliminate, but who's next? If we eliminate ALL SIX candidates at the bottom at once, THEY COMBINED have more votes than any other candidate. Possibly ANY of them might be the second choice of all the others, and that one could rise to FIRST (in an unlikely but not impossible result). So it already seems "unfair" from the first step.
If we break the tie "nicely" like picking cards, one at time, ANY of the 6 might rise, and we'll be playing tie-breakers for another 6-7 round of elimination. And it all seems senseless, since it is so unlikely any of them could win.
BUT if EVEN we play mean, and eliminate ALL SIX, then we're keeping one candidate with only 2 votes, and eliminating six candidates, any of who might gain a second vote on a slower elimination, and again it seems unfair.
Now in a larger election (with more voters), there'll be less ties at the bottom BUT it'll be sort of like the "drawing cards", just a few votes difference will be deciding elimination order, and a slightly different order could perhaps help a different candidate rise.
So seeing all this "protectionism" of bottom-up elimination, playing small differences that are not statistically worth anything. (It's just as unlikely that a candidate with 9.9% of the vote will lose in the end as a candidate with 10.1%, so why do we care to say one is worthy and the other is not?)
This line of reasoning to me supports the idea the LESS elimination rounds is better. How many less? Well, turns out that TWO ROUNDS is simplest. The first round everyone gets an honest vote for their favorite. The second round a majority gets to confirm a majority winner among the final two candidates. In fact, that's what we do NOW in the top-two primary!
So my ANTI-IRV position suggests if IRV is implemented, we're better off skipping the statistically arbitrary "bottom up" elimination and go to a top-two elimination, just like we do now with our top-two primary!
AH, but I know top-two isn't wonderful. It is merely "fair" to the top-two candidates who have "earned" the right to compete head-to-head against each other. BUT what about that "extremist" case I started with? What if there's one "sensible center" candidate who gets squeezed out of the runoff while actually being many people's second choice from both sides?
Well I know a solution. I like to say that IRV is "too careful" at the beginning (caring about elimination order on very weak candidates), and "too arbitrary" at the end, failing to seriously consider 3 strong candidates equally.
How do we do that? It's call Condorcet's method. Condorcet was a French mathematician a few hundred years ago. He said a majority preference can be determine by comparing all candidates head-to-head, and if one choice beats all others, a winner is found.
Interestingly the Rank-preference ballot used in IRV can also be used in Condorcet's method. Every pair of candidates can be compared by looking at the order of choices on each ballot.
What's the problem with Condorcet's method? The primary problem is there's a funny possibility of cyclic preferences. Like the rock-paper-scissors game, it's possible that each candidate sometimes wins, sometimes loses. This is rather unlikely sort of collective opinion, but not impossible if candidates group up on different issues, rather than a usually assumed linear spectrum. Anyway, it is a "problem" and when it happens some sort of tie-breaking rule is needed. Some solutions use "win margins" or other such measures. I prefer to say the highest set of mutually defeatable candidates are in a "real tie" and deserve a random method to break it.
There's a second problem with Condorcet, or at least from the point of view of "plurality" that we use now. There's no measure or rewards for voters who "work together" before the election to get the most first choice votes. It is possible "sensible center" candidate might get ZERO first choice voices and MANY second choice votes, and WIN! Some people don't like that possibility, and I'm with them in that I see candidates who don't get first choice votes must not be very well known or SOMETHING against them, so I worry that perhaps they might win MERELY for being somewhat unknown and merely less openly offensive than another.
For that reason, I don't support a pure Condorcet method, but think it can be used at the END of a runoff process, when all the remaining candidates are strong.
For example, I'd support something like this:
1) Hold a plurality election round (one vote per person), and all candidates above 20% PASS, and those below are removed from competition. (Alternately rank ballots could be used and IRV process until all remaining candidates are above 20%.)
2) If no majority candidate is found from round 1, hold a rank preference election among the remaining candidates to determine a pairwise majority winner.
3) If no majority winner is found in round 2, determine the top set of mutually defeatable candidates from that round, and pick winner among them by lot.
The first round forces voters to work together, identify common interests and reduces the set of candidates with a chance to win. Candidates who are polling far below 20% are encouraged to drop out before the election and supporters can move their vote. (Using IRV in the first round does this drop-out process automatically.)
The second round takes all candidates with a significant level of support, and then treats them completely equally. A standard runoff process can only treat TWO candidates equally, while Condorcet (with the 20% rule) means up to 4 candidates will remain for voter consideration.
THIS approach, unlike runoffs TRULY allows more than 2 parties to exist as equals, and breaks the strangle hold of voters that we have now.
Why not just skip round 1 and let EVERYONE compete in the Condorcet election. WELL, our current politics works by voters talking BEFORE the election and this process helps reduce the set of candidates. A pure Condorcet election needs no selection, BUT it does need very devoted voters since they have to make an informed choice among EVERY PAIR of candidates, and between time and lack of information, voters may be overwhelmed.
Now I'm not claiming politics is ready for a Plurality/IRV-Condorcet hybrid election method. I merely say it has the properties I'm looking for in an election.
BUT my hope lies in the fact that a top-two runoff IS the simplest hybrid option. Rather than a 20% rule, it says top-two remain, and that makes the Condorcet round trivial, and no worry about cyclic preferences. AND someday if we want to go beyond TOP-TWO we have Condorcet's solution. And until then, I say top-two is most fair because it rewards candidates now as expected - by those with the most core support.
MY WORSE FEAR on (bottom up) IRV is an election like this:
Round 1: Republican 40%, Centrist 28%, Democrat 22%, Green 10%
Eliminate Green.
Round 2: Republican 40%, Democrat 32%, Centrist 28%
Eliminate Centrist
Round 2: Republican 52%, Democrat 48%
Republican wins.
BUT a top-two result (IRV or in a top-two primary) might be:
Round 1: Republican 40%, Centrist 28%, Democrat 22%, Green 10%
Eliminate Green and Democrat.
Round 2: Centrist 55% Republican 45%
In a case like this the CENTER-LEFT held the majority, but RIGHT wins merely because the Center gets eliminated without getting a chance to compete head-to-head against his two rivals.
Condorcet (or a runoff hybrid) would allow the Centrist his fair shot in the final round.
I'm all for doing a FULL Condorcet comparison R. vs C., R. vs D., C. vs D. AND that is most fair since the majority winner will be clear. AND if there's a crazy cycle, then its a worthy tie, to be broken randomly.
A top-two runoff SOMETIMES might help a center candidate, sometimes NOT, BUT there's no argument that it rewards power squarely on "first choice" votes. It encourages voters who want to win to find common ground before the election, and rewards those who are best at compromising.
A bottom-up Runoff of IRV isn't BAD, but is also isn't fair to voters who worked so hard to compromise early.
So if IRV is implemented and challenged someday AFTER an election, I'm betting it'll be someday when a candidate reaches SECOND in the first round, and is then eliminated by a close third choice without getting an expected shot head-to-head against either rival.
Whenever elimination order changes WHO gets to compete in the final round, there's going to be sore losers. I just don't wanto fight them, and I'll support their cause - voters and candidates are better served with a top-two elimination process in IRV.
http://www.startribune.com/587/story/1433875.html Minneapolis voters have approved a plan for instant-runoff voting, but the state constitution may not allow it, according to a new attorney general's opinion. The opinion, obtained by the Star Tribune, doesn't explicitly say the Minneapolis system of ranking candidates in order of preference is unconstitutional.
But it concludes that if the closest case to a precedent is followed, instant-runoff voting probably isn't permitted by the state constitution. ... Opponents have threatened a legal challenge, citing a 1915 Minnesota Supreme Court case that struck down a Duluth system that also ranked voters.
The 1915 ruling was against a different rank ballot system called Bucklin which allowed second choice votes to count along with first choice votes, but IRV doesn't allow this. It only counts second choice votes if the first choice is eliminated (just like any runoff process).
If the courts must be involved, I suppose its good to be early, and perhaps a ruling in favor of IRV will reduce the basis for any future rejections of IRV.
The interesting thing for me is, IF something about IRV is found unconstitutional, AND IRV is logically equivalent to a runoff (or top-two primary), then the same unconstitutionality must apply to runoffs or primaries which take candidates off the ballot.
I don't see a constitutional issue with IRV, but there are complaints I take a little more seriously. Specifically, the theoretical possibility that in IRV it is possible to have your candidate LOSE by voting FOR him, and WIN by voting against him. This illogic in election system theory is called nonmonotonicity.
It is possible for the same reason that ANY runoff method allows possible strategic voting. Basically consider a "3-way race". If your favorite is on top, perhaps with 40% support, but needs 50% to win. Then you have two strong rivals, say both around 30% support (being close makes the manipulation possible). Perhaps one rival is "more centrist" and the other is "more extreme" on the opposite political spetrum. You might think your candidate can beat the extreme candidate in the final round, BUT more uncertain against the centrist who will pick up more votes when the extreme candidate is eliminated. SO if a fraction of supporters of your candidate strategically move your votes to the opposite extreme, you can perhaps help the centrist get eliminated.
Interestingly a REAL runoff makes this manipulation easier, and it works in a top-two primary similarly. I can vote in the primary for the opposite extreme, get him into the general election, and then vote my TRUE preference in the general election to beat the weaker competitor.
On the other hand, IRV makes this manipulation more difficult, since you have a rank ballot representing your choices, and you can NOT CHANGE your vote after your stronger rival is beaten. This strategy can still work in IRV more weakly, and only if you trust enough of the centrists supporters will support your favorite second, but the strategy can backfire obviously.
Overall such manipulations are difficult, and require careful analysis of how you expect other voters to act, but I've seen such efforts made, in "party endorsement elections" where delegates vote 10 rounds or more trying to get 60% behind one candidates. Some delegates like to "play" and move their vote around, faking and returning, trying to pretend there's momentum in a direction you like (helping an extremist beat your stronger competitor for instance.) The games happen, whether they actually change anything who knows?
Runoff processes are pretty boring places to play games, and they are mostly a game of dominoes - knowing pretty much the choices will fall in order from the first vote, votes transfering up to stronger candidates in a pretty even and expected way.
The problem I have with IRV is the same problem I have with any runoff process. The hardest to defend aspect of IRV is the "bottom up" elimination, sort of like the endorsement elections, but its just rather arbitrary.
For instance, I've held IRV elections at work for our Halloween constume contests. It is instructive in the difficulties of ties, even if less likely in a larger election. The first round votes were something like this: 5,4,2,1,1,1,1,1,1. There were 17 candidates and 17 voters, so a tough election. Eight voters got zero first rank choices at all, so they are easy to eliminate, but who's next? If we eliminate ALL SIX candidates at the bottom at once, THEY COMBINED have more votes than any other candidate. Possibly ANY of them might be the second choice of all the others, and that one could rise to FIRST (in an unlikely but not impossible result). So it already seems "unfair" from the first step.
If we break the tie "nicely" like picking cards, one at time, ANY of the 6 might rise, and we'll be playing tie-breakers for another 6-7 round of elimination. And it all seems senseless, since it is so unlikely any of them could win.
BUT if EVEN we play mean, and eliminate ALL SIX, then we're keeping one candidate with only 2 votes, and eliminating six candidates, any of who might gain a second vote on a slower elimination, and again it seems unfair.
Now in a larger election (with more voters), there'll be less ties at the bottom BUT it'll be sort of like the "drawing cards", just a few votes difference will be deciding elimination order, and a slightly different order could perhaps help a different candidate rise.
So seeing all this "protectionism" of bottom-up elimination, playing small differences that are not statistically worth anything. (It's just as unlikely that a candidate with 9.9% of the vote will lose in the end as a candidate with 10.1%, so why do we care to say one is worthy and the other is not?)
This line of reasoning to me supports the idea the LESS elimination rounds is better. How many less? Well, turns out that TWO ROUNDS is simplest. The first round everyone gets an honest vote for their favorite. The second round a majority gets to confirm a majority winner among the final two candidates. In fact, that's what we do NOW in the top-two primary!
So my ANTI-IRV position suggests if IRV is implemented, we're better off skipping the statistically arbitrary "bottom up" elimination and go to a top-two elimination, just like we do now with our top-two primary!
AH, but I know top-two isn't wonderful. It is merely "fair" to the top-two candidates who have "earned" the right to compete head-to-head against each other. BUT what about that "extremist" case I started with? What if there's one "sensible center" candidate who gets squeezed out of the runoff while actually being many people's second choice from both sides?
Well I know a solution. I like to say that IRV is "too careful" at the beginning (caring about elimination order on very weak candidates), and "too arbitrary" at the end, failing to seriously consider 3 strong candidates equally.
How do we do that? It's call Condorcet's method. Condorcet was a French mathematician a few hundred years ago. He said a majority preference can be determine by comparing all candidates head-to-head, and if one choice beats all others, a winner is found.
Interestingly the Rank-preference ballot used in IRV can also be used in Condorcet's method. Every pair of candidates can be compared by looking at the order of choices on each ballot.
What's the problem with Condorcet's method? The primary problem is there's a funny possibility of cyclic preferences. Like the rock-paper-scissors game, it's possible that each candidate sometimes wins, sometimes loses. This is rather unlikely sort of collective opinion, but not impossible if candidates group up on different issues, rather than a usually assumed linear spectrum. Anyway, it is a "problem" and when it happens some sort of tie-breaking rule is needed. Some solutions use "win margins" or other such measures. I prefer to say the highest set of mutually defeatable candidates are in a "real tie" and deserve a random method to break it.
There's a second problem with Condorcet, or at least from the point of view of "plurality" that we use now. There's no measure or rewards for voters who "work together" before the election to get the most first choice votes. It is possible "sensible center" candidate might get ZERO first choice voices and MANY second choice votes, and WIN! Some people don't like that possibility, and I'm with them in that I see candidates who don't get first choice votes must not be very well known or SOMETHING against them, so I worry that perhaps they might win MERELY for being somewhat unknown and merely less openly offensive than another.
For that reason, I don't support a pure Condorcet method, but think it can be used at the END of a runoff process, when all the remaining candidates are strong.
For example, I'd support something like this:
1) Hold a plurality election round (one vote per person), and all candidates above 20% PASS, and those below are removed from competition. (Alternately rank ballots could be used and IRV process until all remaining candidates are above 20%.)
2) If no majority candidate is found from round 1, hold a rank preference election among the remaining candidates to determine a pairwise majority winner.
3) If no majority winner is found in round 2, determine the top set of mutually defeatable candidates from that round, and pick winner among them by lot.
The first round forces voters to work together, identify common interests and reduces the set of candidates with a chance to win. Candidates who are polling far below 20% are encouraged to drop out before the election and supporters can move their vote. (Using IRV in the first round does this drop-out process automatically.)
The second round takes all candidates with a significant level of support, and then treats them completely equally. A standard runoff process can only treat TWO candidates equally, while Condorcet (with the 20% rule) means up to 4 candidates will remain for voter consideration.
THIS approach, unlike runoffs TRULY allows more than 2 parties to exist as equals, and breaks the strangle hold of voters that we have now.
Why not just skip round 1 and let EVERYONE compete in the Condorcet election. WELL, our current politics works by voters talking BEFORE the election and this process helps reduce the set of candidates. A pure Condorcet election needs no selection, BUT it does need very devoted voters since they have to make an informed choice among EVERY PAIR of candidates, and between time and lack of information, voters may be overwhelmed.
Now I'm not claiming politics is ready for a Plurality/IRV-Condorcet hybrid election method. I merely say it has the properties I'm looking for in an election.
BUT my hope lies in the fact that a top-two runoff IS the simplest hybrid option. Rather than a 20% rule, it says top-two remain, and that makes the Condorcet round trivial, and no worry about cyclic preferences. AND someday if we want to go beyond TOP-TWO we have Condorcet's solution. And until then, I say top-two is most fair because it rewards candidates now as expected - by those with the most core support.
MY WORSE FEAR on (bottom up) IRV is an election like this:
Round 1: Republican 40%, Centrist 28%, Democrat 22%, Green 10%
Eliminate Green.
Round 2: Republican 40%, Democrat 32%, Centrist 28%
Eliminate Centrist
Round 2: Republican 52%, Democrat 48%
Republican wins.
BUT a top-two result (IRV or in a top-two primary) might be:
Round 1: Republican 40%, Centrist 28%, Democrat 22%, Green 10%
Eliminate Green and Democrat.
Round 2: Centrist 55% Republican 45%
In a case like this the CENTER-LEFT held the majority, but RIGHT wins merely because the Center gets eliminated without getting a chance to compete head-to-head against his two rivals.
Condorcet (or a runoff hybrid) would allow the Centrist his fair shot in the final round.
I'm all for doing a FULL Condorcet comparison R. vs C., R. vs D., C. vs D. AND that is most fair since the majority winner will be clear. AND if there's a crazy cycle, then its a worthy tie, to be broken randomly.
A top-two runoff SOMETIMES might help a center candidate, sometimes NOT, BUT there's no argument that it rewards power squarely on "first choice" votes. It encourages voters who want to win to find common ground before the election, and rewards those who are best at compromising.
A bottom-up Runoff of IRV isn't BAD, but is also isn't fair to voters who worked so hard to compromise early.
So if IRV is implemented and challenged someday AFTER an election, I'm betting it'll be someday when a candidate reaches SECOND in the first round, and is then eliminated by a close third choice without getting an expected shot head-to-head against either rival.
Whenever elimination order changes WHO gets to compete in the final round, there's going to be sore losers. I just don't wanto fight them, and I'll support their cause - voters and candidates are better served with a top-two elimination process in IRV.
posted by Ares Olympus at 7:06 PM
0 Comments:
Post a Comment
<< Home