The Political and Electoral Reform working group, a part of the Occupy Wall Street movement, performed an exit-poll survey experiment on alternative voting methods. It was conducted at a very Democratic-leaning precinct, but the results are still quite informative. By all three methods tested (approval, score, and instant runoff voting), the Green party won, and only under IRV did the Democrats even come in second. In the actual election, the precinct went heavily to the Democrats, of course.
Subscribe to:
Post Comments (Atom)
But all of the alternatives essentially tied..., which suggests going with the one with the marketing first-mover advantage, taking the path of least resistance or highest likely return on our limited political capital, not letting a thousand flowers bloom as they suggest!
ReplyDeletedlw
I wonder if it is more advantageous to have an "all of the above" policy when voting for a single winner. Since all voting methods have pathologies associated with them that could encourage tactical voting, having multiple counting methods that offset each other’s tactical shortfalls would encourage honest voting overall. For example, range voting violates the later-no-harm principle, but not the favorite betrayal criterion. For instant runoff voting, it's vice versa. Both range and IRV, though, fail the Condorcet criterion.
ReplyDeleteI would advocate using range, IRV, and a Condorcet method in the same election. The Condorcet method to use would be a method that I thought up called ReverseIRV, which uses the IRV method after reversing the rankings to elect the worst candidate. That candidate would be erased from the rankings and a new ReverseIRV "winner" would be calculated until one candidate is left. This method would have elected Montroll instead of Kiss in the 2009 Burlington, VT mayoral election.
The range vote would have to be subject to a no tie ranking rule since IRV cannot use tie rankings, but other than that, the votes can be tabulated at the same time.
The only question is how to determine the ultimate winner after all the individual election methods have determined their winner. If all 3 methods return the same winner, then obviously that person should win. If the methods came up with two winners (Candidate A wins under two methods vs. Candidate B with the other method), then the rankings should be scrubbed of all the other candidates and re-run to get a winner between A and B. This would be like looking at the Condorcet matrix, but ignoring all the other table values except A vs. B totals.
The most interesting resolution mechanism would be if the methods each returned a different candidate. I would suspect this would be very rare, but a rare event still needs a contingency measure. I would suggest that approval be used to break the tie. There are two ways to do this. First, the candidate with the highest approval score could be declared the winner. Or second, the candidate with the lowest approval score can be eliminated, and the method used to choose between two candidates (head-to-head Condorcet) can be used with the remaining two.
Is this a complicated way to determine a winner? Yes. But the counter-balancing nature of using multiple methods would force voters to be honest, as any attempt at tactical voting would be thwarted by the threat of harming one's preferred outcome in another method. Plus, it would just be too confusing to try to vote tactically without some advanced mathematics and polling data.
Marketing this method to voters might be hard, but it could easily be sold by just being honest about the ranked ballot's shortcomings. Here's what to tell them:
1. All election reformers consider plurality/first past the post voting to be awful when there are more than two candidates and would replace it with a ranked ballot method.
2. There are numerous ranked ballot methods, and while they are all better than the current system, they each have weaknesses when it comes to tactical voting and the potential to elect the wrong candidate even with honest votes.
3. Using multiple methods will minimize the value of tactical voting and minimize the chance that a spoiler candidate will turn the election results upside down because each method's weakness is not found in at least one of the other methods.
4. It will ensure that the candidate with the broadest support is elected while allowing all voters to vote their conscience.
It's an interesting question to ask "When voting methods disagree, why?"
ReplyDeleteFor instance, IRV disagree with Condorcet methods when the Condorcet winner is 3rd (or worse) in first-place votes, commonly associated with a "center squeeze".
Plurality disagrees with Condorcet when the Condorcet winner is 2nd (or worse) in first-place votes, including center squeeze problems put also an "outside squeeze" AKA simple vote-splitting.
Approval and Condorcet disagree when... well, with certain assumptions, they don't. Unless voters are strategic, and then the true Condorcet winner is more likely to win under approval than under any "actual" Condorcet method.
Like I said, an interesting question.
I think since marketing matters and we are currently in a first-past-the-post system, there can only be one dominant single-winner alternative to FPTP at a time and that precludes the use of multiple methods simultaneously.
ReplyDeleteIt's an interesting thought experiment, but the difficult spade work FairVote has undergone with teaching people about IRV, and the difficulties Approval/Range advocates have had with online petitions, suggests the need to keep it simple and to pick our battles with our current two-party dominated system.
iow, the incidence of whose voters get coerced to vote tactically matters...
ReplyDeleteAn off-center major party can remedy how its supporters would get encouraged to "abandon ship" by moving towards the center. A third party cannot remedy how its supporters are encouraged to abandon it.
IRV may not end 2-party domination, but it helps third parties to have a chance to become one of the local major parties or an important faction within a reorganized major party.
dlw
“It's an interesting thought experiment, but the difficult spade work FairVote has undergone with teaching people about IRV, and the difficulties Approval/Range advocates have had with online petitions, suggests the need to keep it simple and to pick our battles with our current two-party dominated system.”
ReplyDeleteIt’s not simply a thought experiment because I’m trying to deal with the reality of honest voting producing adverse results and the fact that choosing any one method opens up the possibility of voters not voting their true preferences based on the known weakness(es) of the method chosen. We can see the results of both of these problems looking at the IRV results of the 2009 Burlington, VT mayoral race. Burlington shows how a voting system can get it wrong with honest voting, to the point that people become fed up with the voting system and repeal it. If they didn’t repeal it, the city’s GOP voters would eventually be forced to strategically vote for the Democrat to avoid electing a Progressive candidate (which GOP voters in 2009 voted against 3-1).
The holy grail of voting reform is to elect the correct candidate with completely honest votes, with the most emphasis placed on allowing voters to express their true first preference without fear of electing a less preferred candidate. Sometimes the correct candidate can win with a significant number of dishonest votes, such as in plurality when enough supporters of third parties hold their noses to vote for the major party candidate and not spoil the election. But this is not what we are aiming for.
I actually like range voting, but, despite all the Bayesian regret simulations, it is susceptible to the later-no-harm principle and its vulnerability on this has not been tested over the course of numerous elections. What if the election polls are coming out as candidate A having a score of 71.3 and candidate B having a score of 70.2, with other candidates far behind? Wouldn’t supporters of candidate A and B find it in their interest to downgrade the other candidate in an effort to drag their score down, regardless of how they really felt about them? It would seem that voters in a range election are in some sort of prisoner’s dilemma game whereby if they both defect in large enough numbers, an even lesser-preferred candidate could be elected by accident. This is not to say that range is bad; rather, it is to say that range has a real possibility of going bad, not unlike the 2009 Burlington IRV election.
This is why I favor a multi-method system of some kind. If voting tactically caused a win in one method, but had no effect in the others or backfired, then that candidate would still have to face off against the winner(s) of the other methods, so we wouldn’t have a candidate that snuck away a victory because they knew how to game the vote. Likewise, it would be very rare if honest votes failed to deliver the correct winner in at least one of the three methods, so we should be supremely confident that the correct winner would at least make it to the final resolution round regardless of any election pathology. The whole point of the multi-method approach is that a true “beats all” candidate will not fall through the cracks, unlike if we insisted on using one method, despite knowing it had flaws. This is very important, because being saddled with a manipulable and deficient voting system will result in sub-optimal outcomes that will affect the welfare of large numbers of people.
Quite a lot to take in there.
ReplyDeleteFirst, on strategy. If voters wanted to be maximally-strategic (and experience with plurality suggests most of them do most of the time) then yes, if two candidates are doing well, voters would seek to exaggerate the differences in the scores they give them. As I discuss in that link, a very good strategy to follow with score is:
1.) Figure out who the top two candidates are.
2.) Rank the one of those you like better (and any candidates you like more) at max-score.
3.) rank the you like less (and any you like less) at min-score.
4.) If there are any candidates left in between, mark them somewhere in between.
However, there is no prisoner's dilemma here; if everyone follows this strategy, it's not going to allow a truly poor candidate (call them F) to sneak in and win. All those strategic A- and B-loving voters are going to put F at min-score (and even the A-and-B-haters would be scoring one of them at max.) It's disingenuous of you to grant to some voters, but not the rest, enough cleverness to realize that they must exaggerate on A and B, and also insufficient cleverness for them to realize that they must exaggerate about F as well.
Furthermore, if the poll changes the voter's actions, and a second poll is conducted which finds that F truly has a chance of winning (perhaps you're not thinking of a truly poor candidate, just "the opposition") then that changes the voter's actions again, which would then change the polls again, etc. etc., (again, it's disingenuous to grant voters only half a measure of cleverness) but the nice thing about score is that the equilibrium that will be reached is to elect the true/honest Condorcet winner (if one exist, and I think a member of the Smith set if one doesn't); it doesn't handicap the Condorcet winner like IRV does, and won't pick some random schmuck, as can happen in Borda and Condorcet.
(Also, "later no harm" is crap anyway. Voter's don't (and society shouldn't) care about harming candidates, they care about harming themselves; LNH is a petulant "my first choice and no compromise!" demand.)
I've never run into someone who has seriously advocated for "multi-method", as you call it. Be warned, it can get kind of hairy. (Maybe something simple, but robust--like approval?--is best? :)
Logan: It’s not simply a thought experiment because I’m trying to deal with the reality of honest voting producing adverse results
ReplyDeletedlw: The possibility that voting may have unintended consequences for some doesn't mean it's likely or that low-info voters can often rationally predict when such may occur.
Also, if candidates/parties can reposition themselves then the fact that members of a major party who's positioned itself away from the center might be coerced into voting strategically(with IRV) can give their party the right incentive to move to the center.
In a multi-stage game, the "bad" of the non-election of a weak Condorcet winner, as with Burlington, is undone in the next election with learning or re-positioning.
A lot of the problems come from when there are 3 competitive candidates, but this is not a likely, stable situaton imo. The costs of running for a single-winner election, regardless of the rule, make it hard for many serious candidates to coexist.
dlw
For those who can forgive the spam, there's a petition on We the People asking the president to commit to working toward implementing a superior voting system in America.
ReplyDeleteI signed it.
ReplyDeleteI've never been able to use wh.gov, because their system refuses to communicate to email addresses containing a hyphen (and I'm too obstinate to bow to a clearly-incorrect implementation of RFC-819 by using an alternate email!)
ReplyDeleteBut it is a great petition idea.