First of all, congratulations to Michael Brennan on becoming Portland’s first popularly elected mayor in 88 years. Although I do not know him personally, he seems to be a good guy, and I wish him success. It says something that his closest competitors embraced him after the results were finally announced the day after Election Day.
At the end of Election Day, the city of Portland counted 19,583 first-choice votes for mayor using its traditional combination of optical-reader voting machines and poll workers. At that point, Brennan led the field with 5,240, or 26.76 percent of the votes. Ethan Strimling was second with 4,392. Ten of the fifteen candidates in the race got less than 5 percent each.
If you limited the field to just Brennan and Strimling, Brennan had a majority (as one candidate almost always would in a two-way race) and beat Strimling by 8 percentage points, 54 percent to 46 percent.
On the day after the election, in the State of Maine Room at City Hall, the city clerk’s office conducted an instant runoff with the help of outside consultant TrueBallot. The public was welcome to attend and observe the proceedings.
TrueBallot workers scanned 20,212 ballots into their machines. Of those, 578, or 3.5 percent, were invalidated because they were either blank or contained over-votes (ballots on which voters gave the same rank to more than one candidate). By this measure, it does not seem that the 15-by-15, ranked-choice ballot was too complicated or confusing for people to understand.
That left 19,634 valid ballots to be counted in the instant runoff, 51 more votes than the city had counted on Election Day, which seems like a negligible variance.
I compared the totals of the first-choice votes that the city counted for each candidate on Election Day with the totals that Trueballot counted in the instant runoff, and found a 72-vote variance – TrueBallot awarded that many more or fewer votes to the candidates than the city awarded.
For example, TrueBallot awarded Mike Brennan 5,211, or 29 fewer votes than the city gave him on Election Day. It awarded Jill Duson 839, or 5 more votes, than the 834 that the City counted.
That was to be expected. Any time you count the same large number of things twice there is bound to be a difference. All the more so when you use two different systems to do the counting. The magnitude of the variances between the two counts of first choice votes does not strike me as problematic.
But through 15 rounds of instant runoff, the order in which candidates ranked never changed. Brennan finished first after every round, Strimling came in second, Mavadones third, Marshall fourth, and so on.
The first round accurately predicted the final outcome.
It begs the question of whether ranked-choice voting is superior to first-past-the-post voting, or worth the $22,000 that the city paid TrueBallot. My guess is that given a sufficiently large vote, the only time that the order of finish is likely to significantly change in the instant-runoff process is when the field of candidates includes an outlier on the political spectrum. There wasn’t much difference on the issues between the candidates in this election.
Another interesting aspect of RCV is that voters who make high-place finishers their first choice do not participate in the election as much as those whose first choices are eliminated early.
It took 15 rounds of instant runoff to produce a winner. That means that the ballots of people whose favorite was eliminated first could have been counted as many as 15 times, while the ballots of those whose favorite was the eventual third-place finisher were only counted twice, and those of the runner-up once.
That also means the preferences of voters whose candidates get eliminated early can play a significant role in the election, while those of voters whose candidates survive until the late rounds are sidelined.
For example, if supporters of the last-place finisher made the eventual winner their 15th choice, then their minimal preference for the winner would seem to have more impact on the outcome than those whose first choice was for the ultimate runner-up.
But RCV’s big selling point was supposed to be that it produces a winner with a majority and avoids electing someone with a mere plurality. That always had me scratching my head because, in my experience, Portland was not particularly plagued with officials who had been elected with mere pluralities.
The irony is that RCV seems to have created the plurality problem when Portland didn’t have it before.
Maybe it was the novelty of the occasion, but arguably, RCV encouraged marginal candidates to run in the hope that they might attract enough second-choice votes to overtake more mainstream candidates, and thereby made it less likely that someone would achieve a majority in the first round.
The unacknowledged failure is that RCV did not deliver on its promise of a winner with a majority.
Brennan was declared the winner after 15 rounds of instant runoffs with a total of 8,971 votes to Ethan Strimling’s 7,138. But 19,634 valid ballots were cast on Election Day. A simple majority, 50 percent plus one, of that number is 9,818.
Brennan fell 847 votes short.
That’s because 3,525 of the ballots cast on Election Day were exhausted by the time they reached the 15th round: those voters had not made enough selections to be counted by the time their ballots reached that round. They did not care enough for Brennan (or Strimling for that matter) to assign him a rank.
Another way to think of it is that more than a majority of the voters in the instant runoff, 10,663, preferred someone other than Brennan. Over 15 rounds of the runoff process, Brennan’s relative margin of victory over Strimling grew 3.4 percent, from 8 percent to 11.4 percent.
All voting systems are imperfect translations of popular will. Heck, sometimes it’s even difficult to know your own mind. But choices have to be made and officials have to be elected. A voting system defines what it means to be elected. It shapes an official’s legitimacy. In the end, ranked-choice voting was not a debacle. But it cost time and money and it wasn’t particularly superior to the old, first-past-the-post system either.
If we are going to stick with it, RCV might be improved by increasing the number of signatures required to get on the ballot or by imposing some other requirement that limits the size of the field. It might be improved by a provision that broadens the range of candidate viewpoints.
Both goals could be achieved by a return to partisan elections. If you limited the number of the parties to two, you could even be assured that the winner would have a majority.