Maine’s ranked-choice voting system took center stage early Friday, when state officials ran official tabulations that defied projections in every major contest still up for grabs.
State Auditor Matt Dunlap won an upset for the right to face former Gov. Paul LePage in Maine’s 2nd District. Democrats picked Hannah Pingree to be their nominee to succeed outgoing Gov. Janet Mills in a surprisingly convincing fashion. Republicans put Bobby Charles up in that race.
We looked at how the votes flowed and where they diverged from expectations, including our ranked-choice voting simulator that used real first-round results and pre-election polling data.
The biggest surprises came in the 2nd District.
Dunlap, state Sen. Joe Baldacci and former political operative Jordan Wood entered Friday’s count separated by fewer than 2,300 votes out of nearly 79,000 cast. Political newcomer Paige Loud was eliminated first, and her 10% of votes split fairly evenly among the three survivors.
Then Wood was eliminated. Wood’s voters broke for Dunlap by a margin of nearly 18 points to 12 in the final allocation in a split that flipped the race. In the end, Dunlap beat Baldacci 52.5% to 47.5%, overcoming a deficit he had carried since the first ballot was cast and defying national Democrats irked by his primary attempt against outgoing U.S. Rep. Jared Golden.
We had Baldacci winning about 64% of modeled races, with Dunlap and Wood each winning about 18%. While we correctly identified Dunlap as a threat and noted his second-choice support, we underestimated how decisively Wood’s coalition would break his way.
Another factor was the striking 11.5% of Wood voters who ranked neither Dunlap nor Baldacci. This compressed the active vote pool and amplified Dunlap’s margin among those who backed him. In a race this close, that exhaustion rate was decisive.
Pingree won bigger than expected but took a different road.
Going into Friday’s count, BDN simulations had Pingree winning about 30% of modeled races, with Secretary of State Shenna Bellows the narrow leader at 35% and former public health chief Nirav Shah at 25%.
The simulations did not anticipate how wide Pingree’s final margin would be or the road she would take to get there. But give us some credit: We thought Bellows’ path was highly volatile.
She had a good possibility of winning if she could get past former Senate President Troy Jackson after the elimination of last-place candidate Angus King III. But there was always the possibility that she would never make it above Jackson. That’s exactly what happened here.
It mattered because Bellows’ voters broke heavily for Pingree, vaulting her from third place to first. Jackson’s votes in the final round then broke heavily for Pingree. She won 56.2% to runner-up Nirav Shah’s 43.8%, significantly wider than the 4-point victory our poll projected before the election.
Pingree’s broad coalition — from Gov. Janet Mills’ endorsement to her spot in a ranked-choice slate blessed by U.S. Senate candidate Graham Platner — was always her key advantage. But in ranked-choice races, the elimination order can change everything.
The Republican race tracked projections with a major twist.
Charles did more than hold on to win the Republican nomination. He actually built on an 18-point lead in the first round, finishing a rout with 60.3% of votes in the final round to Ben Midgley’s 39.7%. BDN simulations had Charles winning about 76% of modeled races, and that undershot him significantly.
Entrepreneur Jonathan Bush and Midgley were separated by fewer than 300 votes after the first round, and our simulations had them essentially tied in their chance of reaching the final two. The simulations projected Bush outlasting Midgley because he was expected to draw more fourth-place candidate Garrett Mason’s transfers.
Instead, Mason’s votes in round six split nearly evenly between Bush and Midgley — 6.5% to 6.4% — with Midgley edging ahead by the narrowest of margins. Bush was eliminated second to last. By that point, 16.4% of his voters had not ranked either Charles or Midgley candidate, a positive development for the candidate who was ahead.
Even a model that gets the winner right can miss the details that only the count reveals, and we overestimated the number of Republican rankers in our poll.