![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
on voting systems
Two weeks ago, Howard Dean wrote an op-ed in the Times favoring Instant Runoff Voting, and a number of readers weighed in. (I didn't get a letter in within the Times's 7-day window.) Two supported Dean's call for "ranked voting, aka instant runoff voting"; two supported "approval voting, a simpler system that voids [various weird properties of IRV]"; one supported voting under the current system for your favorite candidate, even if that candidate has no chance of winning; one supported proportional representation [although how that applies to a directly-elected President is unclear]; one pointed out that the system isn't going to be changed by vote of people who were elected under the current system; and one makes the peculiar argument that IRV fails "even when the sequential elimination of weaker candidates whittles the number down to two. The reason is that voters who supported weaker candidates can have all their preferred candidates eliminated, so in the end these voters are not counted in the contest between the final two."
I find this last argument peculiar because it's the opposite of an argument I was going to make. Voters whose first choice got knocked out early do have their votes counted in the contest between the final two. The problem is voters whose first choice is one of the final two: nobody ever even looks at their preferences other than the top. Consider an election among Alice Awesome, Peter Prettygood, Oliver OK, and Dr. Evil. If Awesome and Evil are the final two, and Alice is your first choice, it doesn't matter whether you put Dr. Evil in second, third, or fourth place because nobody will ever look at that part of your ballot. IRV pays attention to whom you like, not whom you dislike.
Related to this is another feature of IRV: it tends to favor divisive, polarizing candidates over broadly-acceptable, unifying candidates. Suppose, for example, in the above election Peter Prettygood was the second choice of every single voter in the country, although they divided bitterly between Alice Awesome and Dr. Evil for first choice. The argument could be made that Peter Prettygood is the best candidate to lead the country, because everybody finds him acceptable -- yet Peter Prettygood is the first one knocked out of the race because he's not anybody's first choice.
None of the readers seemed to question the equation of "ranked voting" with "instant runoff voting". In fact, "ranked voting" is about how you cast your vote, while "instant runoff" is one of several possible ways to count ranked votes, the other two leading ones being Borda and Condorcet. Condorcet says "for each pair (A,B) of candidates, how many people prefer A over B?" and declares the election for whichever candidate is preferred over the largest number of other candidates. Borda says "your first-choice candidate gets 4 points, second choice 3 points, third choice 2 points, and fourth choice 1 point" and declares the election for whichever candidate gets the most points. Both of these systems pay attention to all of your ballot, not just your first choice, and both of them would be likely to elect Peter Prettygood in the above scenario.
Of course, both Borda and Condorcet have better mathematical properties, e.g. "if your preferences are the exact opposite of mine, then your vote and mine exactly cancel one another out", which isn't true of IRV or single-vote plurality. But that's perhaps of more interest to mathematicians than the general public.
I find this last argument peculiar because it's the opposite of an argument I was going to make. Voters whose first choice got knocked out early do have their votes counted in the contest between the final two. The problem is voters whose first choice is one of the final two: nobody ever even looks at their preferences other than the top. Consider an election among Alice Awesome, Peter Prettygood, Oliver OK, and Dr. Evil. If Awesome and Evil are the final two, and Alice is your first choice, it doesn't matter whether you put Dr. Evil in second, third, or fourth place because nobody will ever look at that part of your ballot. IRV pays attention to whom you like, not whom you dislike.
Related to this is another feature of IRV: it tends to favor divisive, polarizing candidates over broadly-acceptable, unifying candidates. Suppose, for example, in the above election Peter Prettygood was the second choice of every single voter in the country, although they divided bitterly between Alice Awesome and Dr. Evil for first choice. The argument could be made that Peter Prettygood is the best candidate to lead the country, because everybody finds him acceptable -- yet Peter Prettygood is the first one knocked out of the race because he's not anybody's first choice.
None of the readers seemed to question the equation of "ranked voting" with "instant runoff voting". In fact, "ranked voting" is about how you cast your vote, while "instant runoff" is one of several possible ways to count ranked votes, the other two leading ones being Borda and Condorcet. Condorcet says "for each pair (A,B) of candidates, how many people prefer A over B?" and declares the election for whichever candidate is preferred over the largest number of other candidates. Borda says "your first-choice candidate gets 4 points, second choice 3 points, third choice 2 points, and fourth choice 1 point" and declares the election for whichever candidate gets the most points. Both of these systems pay attention to all of your ballot, not just your first choice, and both of them would be likely to elect Peter Prettygood in the above scenario.
Of course, both Borda and Condorcet have better mathematical properties, e.g. "if your preferences are the exact opposite of mine, then your vote and mine exactly cancel one another out", which isn't true of IRV or single-vote plurality. But that's perhaps of more interest to mathematicians than the general public.