Ranked Choice Voting
Ranked Choice Voting for Boulder’s mayor: The good and the bad
Ranked choice voting (RCV) is a voting system where the voters rank order the candidates using a grid on the ballot that has the candidates’ names down the left side and columns labeled first, second, third, etc. across the top. Voters fill in the ovals in the grid to “rank” order the various candidates.
The problem with RCV is that its structure can arbitrarily drop the candidate with the most support, so we could very easily end up with a mayor who is not the people’s preferred choice. But it does pretty much guarantee some finality. (The only problem is if it ends up with a tie at any point, which is highly unlikely with tens of thousands of voters.)
To show how this sub-optimal outcome might happen in a simple example, imagine the current race but with only three mayoral candidates – Speer, Yates, and Brockett. Suppose all voters think Yates would be a good mayor, and so rank him as their first or second choice. But also suppose that 34% percent of voters (barely over a third) rank Speer first, Yates second, and Brockett third. And 36% of voters rank Brockett first, Yates second, and Speer third. And the remaining 30% of voters rank Yates first, and split on the second and third, with 15% putting Speer second and Brockett third, and the other 15% putting Brockett second and Speer third.
The Analysis:
The way RCV works, the candidate who receives the fewest first place votes is dropped in the first round. Then, for those voters who voted for that candidate first, their first choice is “erased”, their 2nd place choice becomes their first place, and their 3rd place choice becomes their 2nd, and the ballot counting process is repeated, with the next candidate with the fewest first place votes dropped. This is done as many times as needed until only one candidate is left.
In the example above, Yates get dropped in the first round because he got the least first place votes. Then the RCV contest goes to the second round and is between Brockett and Speer: Voters who gave Yates first place in the first round have their ballots reallocated based on their second-place votes. So, Brockett gets 36% plus 15% and Speer gets 34% plus 15%. Thus, Brockett wins with 51% versus Speer at 49%.
The Problem:
But this is not the optimal result. Here’s why: If we were to have the voters compare these candidates in pairs, we would see that 64% (34% plus 30%) prefer Yates to Brockett, and 66% (36% plus 30%) prefer Yates to Speer. In other words: If Yates just went up against Speer, Yates would win. And likewise, if Yates went up against Brockett, Yates would win.
That makes Yates the clear head-to-head winner, and so should become mayor. But under RCV’s arbitrary system, Yates was dropped in the first round, producing a polarized, sub-optimal outcome – exactly what we don’t want.
This bizarre result actually happened in Burlington Vt., after which the town’s voters intelligently repealed RCV.
The Alternatives:
There are much better alternatives to RCV. Perhaps the obvious one is, as in the example above, simply pit each candidate against each of the others on an individual basis. This can be done using the same RCV ballot format. But the votes are scored differently: simply total up the number of ballots in which Yates is ranked higher than Speer and vice versa, Yates is ranked higher than Brockett and vice versa, and Speer is ranked higher than Brockett and vice versa.
This system is called Condorcet; it is well known and studied. It can be done using some straightforward existing software for the ballot counting machines, and with the same simple ballot layout. And, conveniently, it is consistent with the current City Charter language, so, it could be implemented without another ballot vote.
The only time that the head-to-head (Condorcet) system fails is in the highly unlikely case that the candidates are ranked circularly: Voters rank Yates higher than Brockett, Brockett higher than Speer, and Speer higher than Yates. Fortunately, there are various systems that can address this highly unlikely scenario, and they are no more arbitrary (and arguably less so) than the standard RCV approach of dropping the candidate with the least first place votes in the various rounds.
Another alternative system to the standard RCV is called “Approval Voting” (AV). Here where voters get to vote for as many candidates as they “approve” of, and the candidate with the most total votes wins. The ballot is very simple; all the voters have to do is check which candidates they approve of. AV generally leads to a winner who is supported by a majority, and so is less polarizing and more satisfying than RCV.
An AV mayoral vote could be run by the county clerk just as our current City Council elections are, using the same software. The only differences are that voters are not limited to voting for only five candidates (if more than five run for mayor), and there’s only one winner.
All these alternatives should have been considered by the Council before putting RCV on the ballot. Now voters may have to deal with the outcome.