Friday, January 11, 2008
As always, the Final Word is heard 3-5pm on AM1280 the Patriot (click that link and then the Listen Now in the upper lefthand corner) or later via podcast when the podcasts gods choose to post our shows. Like monetary policy, shows are posted with long and variable lags.
Back to IRV. Let me help people think about it by a couple of simple thoughts. Why would IRV be bad? After all, we had three viable candidates in the 1998 gubernatorial race won by Jesse Ventura over Norm Coleman and Skip Humphrey. Humphrey came in third. Suppose we had IRV in the state. How would Humphrey's vote been distributed? Would Coleman have gained enough of those votes to overcome Ventura's three percent plurality. At least in one study, the answer is no. So plurality voting -- the system we use most everywhere in the US -- gave us the same result. (My apologies if you can't read that study -- it's good to work at a university.)
We would say that Ventura was a Condorcet winner -- he would have won in a pairwise vote against either Coleman or Humphrey. And perhaps one reason why Ventura would have won is that he turned out people that would not have voted otherwise; 7% more voters participated in 1998 than 1994, and the study cited above indicates that the Ventura presence could account for just about that size effect. An exit poll taken that day indicates that of those who did vote, Ventura voters preferred Coleman over Humphrey, 56-44.
So IRV would not have made a difference there. Where would it? It could if you ended up with cycling, where in a pairwise contest between each of the three candidates you wouldn't find one that beats the other two in heads-up. And it would be perilous if we found instead that Coleman would beat Ventura and Humphrey would beat Ventura in pairwise voting, but Ventura wins in a three-way, plurality-voting contest. That would seem to be something we would want to avoid and could be a case for IRV. And yet, as we joke a bit about voting for bacon, there is a very serious question whether we might end up with a Condorcet loser in the primaries with some states using winner-take-all allocation of delegates. On the Democratic ticket, let's suppose Sen. Clinton is the plurality leader. She would be a Condorcet loser if she would lose a two-person race against either Sens. Obama or Edwards. (I make no claim that this is so.) I'll let others think about how that applies on the GOP side, but I think it fair to say IRV would make the GOP primary different.
I find it interesting that the claim made by MVA is that IRV is too complex. Indeed, the best case one can make for plurality voting is its simplicity. (Try reading the rules for the DFL's walking caucus sometime. They aren't easy.) There are other good reasons to oppose IRV, though they are a bit harder to explain, having to do with lack of single-peaked preferences and the number of parties in play. I've not worked on electoral theory for awhile so I am not going to try to say I understand every bit of this; I don't. Suffice to say, for example, that it can be shown that you can get different electoral outcomes by changing IRV rules only slightly to drop the candidate with the most last-place votes rather than the fewest first-place votes. This is known as the Coombs rule. You could argue Coombs is a better rule, but it would be no less confusing.
A last thought, and a question I'll ask Andy tomorrow: Suppose we could show empirically that IRV increased voter turnout, since one could vote his preference first and then between the two major party candidates later. Would increased turnout be a good thing, and wouldn't you want to support IRV then if so? I know they argue that it hasn't, but would finding enough counter-examples to theirs be persuasive?