It's distracting because their stats aren't exactly what happened in that election. But, because of history and the names, it seems right that Reagan should win, and just weird that Anderson should win. Especially since we have this history of feeling dominated by the two major parties, and with Anderson being an independent.
So I will redraw the scenario here. Same stats, same candidate make-up. But without the names:
Votes | 1st | 2nd | 3rd |
48 | |||
47 | |||
4 | |||
1 |
That's better. IRV: Red wins. Condorcet: Yellow wins.
This one is interesting because at first glance it certainly does seem as if either red or blue should win. If you look at first-place votes, red and blue are way ahead of yellow. The most damaging argument against Yellow is that 95% of the voters don't even want Yellow as their first choice.
But if we look at the rankings, it's not that simple. Third place here doesn't just mean a good show like in a horse race, it means last. The people that voted red first really don't want blue to win, and vice versa. In fact, we can tell that if either red or blue is voted in, around half of the voters will be very unhappy.
So what is the goal of a vote? Is it to make a horse race with a dramatic finish where you can either win big or be crushed? Or is it to capture the collective consensus of the entire voting population?
If it's consensus, then I believe Yellow is the winner. 100% of the voters ranked Yellow either first or second. No one hates Yellow. 52% of the voters hate Blue. 48% hate Red. Neither would well represent the entire population. And Yellow would. No one voted only for Red (or Blue) with no other choices - all are indicating they'd accept Yellow as a second choice. Yellow reflects the consensus choice of the voters.
Voting systems are eventually dependent upon subjective preferences of
the people advocating them. If the goal is to focus mostly on
plurality-style counting, only looking at preferences if the outcome is
in doubt, and if the goal is to put as much weight as possible on the
1st-choice candidates and risking having your preferences ignored if
support the wrong candidate, then IRV is the right voting system. But
if the goal is to have all your preferences paid attention to in the
counting, and if the goal is to find the consensus choice of the entire
group, Condorcet (pairwise) would be the better choice.
Posted by Curt at November 17, 2002 11:26 PM