Just for the record, I'd like subscribers of this list to be aware of
the 2009 Mayor's election in Burlington, Vermont (USA).
Candidate(Party)..............1st Round.........2nd Round.....Final
Bob KISS(Progr)................2585(29%).........2981.............4313 (wins)
James SIMPSON(Green).....35 (0.4%)
#Voters Their Vote
The election was held using Instant Runoff Voting (IRV), the second time Burlington had used IRV to choose their mayor. The link page gives the tally of the ballots cast and an analysis of five ways that IRV functioned poorly. The biggest complaint was that IRV clearly elected the wrong guy. The Democrat, Montroll, would have won two-person runoffs against all other candidates in the race, i.e. he was the "beats-all" or "Condorcet" winner, but IRV instead elected the Progressive, Kiss, over
Montroll and the Republican, Wright.
Voters in Burlington were so dissatisfied with the result that they
repealed IRV and went back to using the "tried and true" system they were familiar with, Plurality with a top-two runoff. Which is really rather sad, because that system is no better than IRV and often worse. It would even have given the same result as IRV in this case.
This election gives a clear real-world example of what I call the
"center-squeeze effect". The Democrat was the centrist in this race, between the Republican and the Progressive. He got first-choice votes from a slice of the electorate in the center, but the Progressive got a slice from the center and all of the left wing, and the Republican got a slice of the center and all of the right wing. So the centrist came in THIRD in first-choice votes, and was eliminated before the final runoff between the Republican and the Progressive. So Kiss was elected, despite a clear majority of the voters preferring Montroll over Kiss.
The center-squeeze-effect can happen with three candidates, and it
becomes more likely when there are more than three. (Top-two runoff suffers from this problem also, as does simple Plurality.) It is not unlikely or "rare", on the contrary we can expect it to happen often in IRV elections.
The linked page shows that this election ALSO is a case of the plain-old
"spoiler effect". If the Republican, Wright, had not run, Montroll would have won over Kiss. By joining the race, Wright did not win, but he took enough votes away from the centrist to cause the centrist to lose to the
Progressive, a worse result from the point of view of Wright's supporters. So, this example demonstrates that IRV does NOT eliminate the "spoiler effect", as is often claimed.
My conclusion? I think advocates of IRV should seriously consider an
upgrade that eliminates this problem. One recent proposal is called WBS-IRV (Woodhall-Benham-Smith IRV):
1. The voters fill out ranked ballots as in regular IRV, which are
collected and tallied.
2. The administrators of the election check to see if there is one
candidate (of those remaining in the race) who beats each of the others in "instant" two-person runoffs. If so, they win the seat. If not, then
3. The (remaining) candidate with the fewest top-rank votes is
eliminated and their ballots reassigned to the candidate ranked next on each ballot. Go back to step two.
In other words, at each round, instead of checking to see if any
candidate has a majority, they check to see if any candidate is a "beats-all" or "Condorcet" winner.
This version of IRV fulfills all the promises usually made by advocates
of regular IRV. It also does what most IRV advocates are really after, which is to introduce voters to the Single Transferable Vote, the ranked ballot, which can be used for a good form of Proportional Representation in multi-seat races.