AUGUSTA, Maine — Ranked-choice voting might have kept Gov. Paul LePage out of office, but a Bangor Daily News simulation cast some doubt on the impact of the “spoiler” phenomenon that many left of center have blamed for his 2010 and 2014 election wins.
The results paint a muddled picture of the change to elections under Question 5 on the Nov. 8 ballot, which would move Maine to ranked-choice voting in gubernatorial, congressional and legislative elections with three or more candidates, barring constitutional concerns.
In these elections, a winner is declared if a majority picks a candidate as their first choice. But if not, the candidate with the lowest number of first-place votes is eliminated and second-place votes for that candidate are reallocated. The process is repeated until someone wins a majority.
The push comes after nine of Maine’s last 11 gubernatorial elections — including the Republican governor’s wins in 2010 and 2014 — were decided by a plurality and not a majority of voters.
Advocates of ranked-choice voting say it will allow Mainers to pick preferred candidates and minimize the strategic voting that Maine saw when most left-of-center voters rallied behind independent Eliot Cutler in 2010 and Democrat Mike Michaud in 2014, but not enough to prevent a LePage victory.
Strategic voting was still apparent in a BDN survey that put more than 900 people through an online re-run of both elections using ranked-choice voting. A statistical model favored a Cutler win in 2010 and a LePage win in 2014, despite a liberal exodus from Cutler to Michaud.
The BDN’s sample was self-selected, which means that many who signed up to participate are interested in ranked-choice voting, and it was heavily biased toward Democrats, who made up 50 percent of the sample to Republicans’ 8 percent. That means the voters in our sample likely behaved differently than a representative group of the Maine electorate.
To deal with this, we used actual results from 2010 and 2014 and paired them with choices indicated by those in our sample using a statistical technique called bootstrapping. We then simulated both elections 1,000 times with ranked choices. Data Scientist Jake Emerson dives more into the methodology on The Level.
Liberal voters’ rally around Cutler could have kept LePage out of office in 2010, but the governor may have still won in 2014.
LePage’s two elections are a motivation for many supporters of ranked-choice voting. Since his win with nearly 38 percent to Cutler’s 36 percent in 2010, advocates used it as an argument for the method, saying Cutler likely would have won.
During and after 2014’s campaign, Democrats labeled Cutler as a spoiler. However, he was on Fox News on Monday saying he could have won that year if ranked-choice voting were the law, even though he only got 8 percent of votes to LePage’s 48 percent and Michaud’s 43 percent.
But the BDN’s model found that while he would have been favored in 2010 under a ranked-choice system, LePage may have taken the 2014 election.
In an average of 1,000 simulated elections, we found that LePage stays ahead of Cutler as the two longshot independents — Kevin Scott and Shawn Moody — are eliminated. But Cutler wins when Democrat Libby Mitchell is eliminated and 86 percent of her votes go to him.
You can follow this better with the above diagram, which shows how Cutler’s stream of voters merges with Mitchell’s to put him ahead of LePage.
This is much different in 2014, when Michaud mostly consolidates left-of-center voters, yet still loses to LePage by an average of just over 37,000 votes.
You see in the diagram that it’s because our model sees a group of Cutler diehards that are more conservative than the group boosting him in 2010, breaking in LePage’s favor.
Strategic voting against LePage was still apparent in our sample, raising doubts that ranked-choice voting would tamp it down.
Because our model pairs plurality results with ranked choices, it’s fair to argue that we sell Cutler’s 2014 chances short. Exit poll data collected by the Center for Election Science said he would have been favored to win by to edging out Michaud in a first round and LePage in a second round.
We don’t have polling, so we’re not disputing that as a possibility. But our sample was full of examples of the strategic voting that has been at the heart of the criticism over the last two elections, at least casting doubt on the certainty of that outcome.
In our raw sample, 47 percent of Cutler’s voters in 2010 switched four years later to make Michaud their first choice. Virtually all Mitchell voters — 96 percent — went to Michaud in 2014.
Many of these people indicated more agreement with Michaud, negative attitudes toward Cutler’s second campaign or both in email interviews.
— Ryan Woodward, a 31-year-old independent and educator from Bucksport, said he chose Cutler in 2010 “to strategically vote against LePage” but backed Michaud in 2014 “because I truly believed in positive campaign style.”
— Alex Yakovleff, a 32-year-old independent and retail clerk from Portland, said he broke late for Cutler in 2010, but in 2014, he was “pretty resentful that a moderate-liberal was occupying political space that Michaud also occupied” and voted for the Democrat.
— Terran Siladi, a 25-year-old independent and mechanical engineer from Portland, said he was “very excited” about Cutler’s 2010 campaign, but four years later, he found it “much more desperate and less energetic” and backed Michaud.
— An independent Ellsworth lawyer who didn’t want to be identified said in 2010 he “preferred Cutler’s appeal as the next [independent U.S. Sen.] Angus King,” but in 2014, he “seemed to waffle on whether or not he was serious about running, and that turned me off,” so he voted for Michaud.
— Charlene Farnham, 65, a retired teacher and Democrat from Hampden, said she “felt Cutler was just a spoiler” the second time around.
For the record, four of these five people said they liked ranked-choice voting better than plurality voting. But it shows that strategic voting may not go away.
BDN data scientist Jake Emerson designed this report’s statistical model and contributed analysis.