Re: For Real? VT: "My Turn: Math works for instant runoff"
- --- In RangeVoting@yahoogroups.com, "Joyce McCloy" <jmc27106@...> wrote:
>In 1998, Howard Dean (D) defeated Ruth Dwyer (R) 55.6% to 41.1%.
> Election methods experts, math experts, what say you?
> VT: My Turn: Math works for instant runoff
> My Turn: Math works
> for instant runoff
> I am not a candidate nor an office holder. I am
> simply your neighbor. One of the subjects I teach
> covers various nonweighted (each votes counts
> the same) voting methods: plurality, plurality
> with elimination (aka instant runoff, or IRV),
> Borda Count (used for the Heisman Trophy,
> among other applications), Condorcet (aka
> pairwise) and variations on these. The overall
> conclusion of this examination is represented by
> Arrow's Impossibility Theorem, which states that
> no voting system satisfies all fairness criteria.
> For fun and by way of demonstration, our
> textbook has a sample election with preferential
> ballots formulated so that each of these
> methods produces a different winner.
> So it is left to decide which system works best
> for the election being held. For elections to
> office, the two main contenders are plurality
> (most votes wins) and plurality with elimination
> (IRV). Our text states, "In spite of its frequent
> usage, the plurality method has several flaws
> and is generally considered a very poor method
> of choosing the winner of an election among
> several candidates."
> The advantages of IRV are these: First, IRV
> prevents the sinking feeling that one cannot vote
> for one's favorite candidate because splitting an
> ideological vote will actually award the election
> to your least favorite candidate. This is so
> insidious that, in 2000, Howard Dean actually
> admonished voters not to vote for Anthony
> Pollina because that would be a vote for Ruth
> Dwyer. Obviously, most people who would vote
> for Pollina would prefer Dean over Dwyer, but
> many felt they must vote for Dean, strategically.
In 2000, Anthony Pollina (P) also entered the race, and the results was Dean 50.4%, Dwyer 37.9%, and Pollina 9.5%. If we assume that Dean and Dwyer would have done about the same as in 1998, them 5.4% of his support came from Dean and 3.2% from Dwyer.
And even under IRV, Dean's strategy would have been rational. Pollina would have had to finish ahead of Dean to advance make it into the runoff. This wouldn't have happened if he went around claiming that he and the incumbent governor were like tweedledee and tweedledum, mark us one and two or two and one, it doesn't matter.
So he would have had to go on the attack against Dean. This would likely drive some Dean support to Dwyer. And if Pollina then finished ahead of Dean he would have to hope that Dean supporters would transfer to him. Some would not transfer at all, which would be as good as a vote for Dwyer, and others would not switch to Pollina after he had trashed Dean.
So Dean would still be better off convincing voters that their vote was wasted, which avoids any direct attacks on Pollina.
In 2008, Pollina again ran for governor, this time as a independent, and finished second to Jim Douglas 53.4%, Pollina 21.8%, and Gaye Symington (D) 21.7%. Pollina had the endorsement of several unions. The result shows the wisdom of Dean's strategy in 2000.
> IRV eliminates such strategic voting. If you areActually in 2000, Brian Dubie was defeated in the race for Lieutenant Governor.
> an ideological liberal, you could have voted for
> Pollina, knowing that even if he didn't win, your
> vote would then go to Dean, and not to Dwyer.
> IRV thus solves the "Dubie" effect, as Brian Dubie was elected that same
> year as lieutenant
> governor with a mere 38 percent of the vote.
He was elected as Lieutenant Governor in 2002, when he received 41.1% (not 38%) of the vote, against Democrat Peter Shumlin 32.1% and Anthony Pollina with 24.5% of the vote.
> InDubie has been re-elected in 2004, 2006, and 2008 with 56.0%, 51.1%, and 55.0% of the vote.
> case you didn't think it mattered much, this has
> resulted in Dubie (whom 62 percent of voters
> rejected) having a chance of being our next g
> overnor. Once in office in Vermont, most are
> retained, so long as they don't commit a high-
> profile negative act. He hasn't done this and so
> has been re-elected and is still in office.
Maybe the author thinks that Arrow's Theorem has something to do with with selecting the "best" candidate. Shumlin was "rejected" by 68% of the voters, and Pollina was "rejected" by 75% of the voters. Dubie was only "rejected" by 59% of voters.
It appears the author has conflated an election where the Republican gubernatorial candidate lost with 38% of the vote, with an election two years later where the Republican lieutenant-gubernatorial candidate won with 41% of the vote, perhaps because his favored candidate finished 3rd in both elections.
It is fortunate that he is a math teacher rather than a political science teacher.
> It has been often shown that non-instant runoffsThis begs a citation.
> generally attract about one-third of those who
> voted in the original election.
> Also, those withAlternatively, it permits voters to compare the two finalists, head to head. Bob Kiss is universally acknowledged to be a nice guy. That might not necessarily be the most desirable quality in a mayor.
> bigger war chests will have an unfair advantage
> in a delayed runoff. And since delayed runoff is
> basically the same process as instant runoff, it's
> more fair to have all the voters involved, voting
> on the basis of same campaign.
> A delayed runoffThis very rarely happens. If it does, it means that there are 4 or 5 candidates who have quite similar support, and in which one by one elimination is probably no better than mass elimination. These are the type of elections where you can change the counting rules and get a different result in every case.
> is actually less fair, since it immediately
> eliminates all but two candidates, whereas IRV
> eliminates them one at a time, and reconsiders
> the result each time.
> IRV also informs the campaign with betterWhy will it result in better discussion of issues? If the candidates appear to all agree on the issues, or deliberately diminish the differences, on what basis are voters going to rank the candidates? Personality, physical appearance, name recognition, greater advertising?
> discussion of issues and less name-calling.
> Candidates can't throw mud since they mightIt all depends. If the opponent is going to get more first preferences than you, you need to attract some of those, by creating doubt about the other candidate. If he is going to finish behind you, there is no need to attack him, but rather just ignore him.
> need to be a voter's second choice.