There is a tradition among scientific disputants, rarely imitated by political ones, of addressing themselves to the strongest arguments against their own view. In that spirit, I want to point out something about instant runoff voting (IRV) that I do not particularly like.
Suppose that, in our Minnesota governor's race, the Independent in the contest, Tom Horner, who appears to be lagging, were to gain ground and pull just about even by election day, so that the final result was, say: Emmer, 35%; Dayton, 33%; and Horner, 32%. In an IRV contest, this would not make Emmer governor--a good thing, I've argued. Instead, Horner would be eliminated from contention, and the ballots cast for him would be redistributed among Emmer and Dayton, according to the candidate named on these ballots as second favorite. Presumably this would put someone--I've speculated that it would be Dayton--over the 50% threshold needed to win election.
That would in my opinion be a fine outcome but, by the logic of IRV, is it really best? It seems very likely that nearly a hundred percent of all voters named Horner as either their first or second choice, yet he is the one who is eliminated. Meanwhile Dayton, regarded as anathema by 35%--that is, by all the Emmer voters--takes office. Notice what happens if Horner and Dayton switch places in my speculative results. Now the outcome of the first round of counting is: Emmer, 35%; Horner, 33%; and Dayton, 32%. Dayton therefore is eliminated, his ballots are redistributed according to the indicated second choice, and Horner wins.
The final outcome hinges on the nearly dead heat contest for second place in the first round of counting. Only the second choices of voters who prefer the third place finisher matter. In the scenario in which Dayton becomes governor, Horner gets no credit for being the second choice of nearly all the Emmer voters.
Well, when you are in the laboratory you can cook up any scenario you desire. The question is not which system is perfect but which is best. I still think that IRV, for the reasons given here, is superior to what we now have.
Comments