Qaanol wrote:For an even more striking example of why the majority criterion is wrong, consider a 3-person race between A, B, and C. The voters fall into two groups.
51% of voters rate the candidates like this: A = 100, B = 90, C = 0
49% of voters rate the candidates like this: A = 0, B = 90, C = 100
Now in a straight majority choice, candidate A will take 51% of the vote and candidate C will take 49%, while candidate B will get zero votes. However it should be obvious that B is far and away the best choice to represent the voters. Yet under Plurality, or Majority Judgment, or your Bucklin variant, candidate B will not receive a single vote. Any voting system that can reasonably be expected to take the best candidate, and wind up with that person getting no votes, is seriously flawed.
If you have strategic voters, range voting is not perfect here either. In an election with all-strategic voters, A wins this election. Provided some polling suggests (and it will) that A is more popular than C, but B is broadly well liked by supporters of both A and C, then A voters should rank B as 0. Range voting is highly susceptible to strategic voting, and indeed, any voter should rank all except their number 1 choice at 0.
Your last sentence here is objectively and unambiguously wrong. Take the simple case of 2-range voting, aka Approval Voting. The best strategy for a rational voter is to calculate, based on pre-election polling or whatever other method, how likely each candidate is to win, and use that along with their internal preference for each candidate to determine the expected value of the election for them based on the information they have. Then vote for all candidate they like better than than expected outcome.
Your claim amounts to, “In Approval Voting, people who rate Ralph Nader at 100, Al Gore at 90, and George W. Bush at 0, should vote only for Ralph Nader.” That is clearly false, because even in plurality where you are forced to vote for exactly one candidate, the best and most common strategy for those voters was to vote for Al Gore. Under Approval Voting, it is foolish to think those voters would suddenly abandon Al Gore, and much more reasonable to expect they would approve of both Nader and Gore.
Furthermore, with the election as I described it, with the two groups being so close in number that it is essentially a coin-flip to guess which group will have higher actual turn-out, any given voter would be a fool to disapprove of B. This voter thinks B is great, certainly much better than that awful opposing candidate, and would be entirely happy for B to win.
In other words, polling will not show that A is significantly more popular than C, because that is not the case. They are almost evenly split, and on any given day the numbers could jump several points one way or the other based on recent events, or just who feels motivated to actually vote. That is to say, C is just as likely as A to win in a plurality vote, as far as anyone can predict.
omgryebread wrote:Thesh's criticism of range voting is also valid: it assumes approval of a candidate can be quantified, and that all voters will use the same metric for quantifying it. And I agree that the situation seems very unlikely.
Your argument here is a strawman, even though you may not realize it. Yes, you are absolutely correct that individuals’ internal rating for voters are not on the same scale as each other, and that there is no voting system that will accurately measure it. The point is, even given all those short-comings
accounting for the fact that real voters exaggerate there ratings at least enough to put their favorite at 1.0 and their least favorite at 0.0, range voting still
does the best job of electing the candidate who best represents the collective will of the people.
That is to say, an omniscient being who could magically add up all the voters’ true ratings for each candidate using a uniform scale, would be able to say which candidate has the highest level of voter support. Then, looking at the results of thousands and thousands of elections, that being would be able to say, on average, how much worse of candidates are elected by Range Voting than the magical best winner would be, in terms of total voter support. And the being could do the same for Approval Voting, and for Plurality, and IRV, and Borda, and Condorcet, and every other election system. And then the being could compare the results, and say how well each voting system does in terms of electing the most-supported candidates.
And when you run computer simulations where you get to be that omniscient being, and the voters get to behave in their own best interest to an extent programmed by you, it turns out that Range Voting and Approval Voting are very close together at the top of the list. The other voting systems are much further back, especially when the voters are highly strategic. And when the voters are highly strategic, Range Voting has almost no improvement over Approval Voting anyway.
So Approval Voting is among the very best election systems for picking the most-preferred candidate, and it is extremely simple to use.
omgryebread wrote:Voters don't tend to rank politicians on a scale of "they agree with me on X positions, disagree with my on Y" and use that to calculate their approval. My agreement with Johnson on drug policy doesn't come anywhere close in importance to my disagreement on health care. In fact, his position on health care would cause me to honestly (not strategically) give him a 0, or perhaps a 1 or 2 to put my preference of him over Romney. Many voters would 0 a politician who supports abortion.
Are you intentionally contradicting yourself here? You start by saying “Voters do not do XYZ”. And then you say, “I would do XYZ, and many other people would do XYZ”. In fact, you seem to be saying that yes, voters do
often base their opinions of politicians based on issue positions. There are other factors besides traditional “issues”, to be sure, but the fact remains that, by whatever means a voter comes to have an opinion of a candidate, voters do in fact have opinions on candidates.
omgryebread wrote:If A and C voters agree that strongly, there's clearly some very big differences between the two. I don't see how a B could exist that straddles the divide that well.
It was an example with intentionally-exaggerated numbers. Change the ratings to “A=70, B=65, C=40” for group 1, and “A=40, B=65, C=70” for group 2, still with a 51-49 split, meaning it’s so close that going into the polling booths you don’t have any way to know which group really will have more voters turn out.