Simulated voters
4,597
Synthetic Poll · U.S. Senate — Maine 2026
Synthetic Poll · If Graham Platner exits
We simulated Maine’s Senate race nine times — once for each Democrat who might replace Graham Platner. Two of them beat Susan Collins.
No phone banks, no field time: a simulated panel of 4,597 real registered Maine voters, drawn from the national voter file across all 16 counties, each asked how likely they are to back the Democrat in every matchup. The readings are validated against five certified past Maine elections and anchored to the current Collins–Platner public polling. The short version: a strong replacement keeps the seat competitive — and the ranking below is the firm signal.
Counties covered
All 16
Margin of error
±4.1 pts
Validated against
5 certified races
The bottom line
Every replacement against Collins
Finding
Heather Cox Richardson runs strongest against Collins (Richardson +7); Valli Geiger runs weakest (Collins +2). A strong replacement keeps the seat competitive.
Each potential Democratic nominee was tested head‑to‑head against Susan Collins on the same voter panel, with each matchup asked on its own so question order couldn’t sway the answers. Read the ranking as the firm signal; any two candidates within about 4 points — the margin of error* — are a tie.
| Democrat | Result vs. Collins | Read |
|---|---|---|
| Heather Cox RichardsonHistorian, Letters from an American | Richardson +7 | Beats Collins |
| Shenna BellowsMaine secretary of state | Bellows +5 | Beats Collins |
| Jared GoldenU.S. representative, ME‑02 | Golden +2 | Toss‑up |
| Troy JacksonFormer Maine Senate president | Jackson +2 | Toss‑up |
| Patrick DempseyActor, Maine native | Dempsey +1 | Toss‑up |
| Nirav ShahFormer Maine CDC director | Shah +1 | Toss‑up |
| Dan KlebanMaine Beer Company co‑founder | Collins +1 | Toss‑up |
| Jordan WoodFormer congressional chief of staff | Collins +2 | Toss‑up |
| Valli GeigerState representative, Rockland | Collins +2 | Toss‑up |
* This is not a textbook sampling figure. It is measured directly from the method’s track record: after the correction, we compared its reading of the five certified past Maine elections to the actual results, and the typical miss was about 4.1 points. Read it as: any two candidates within about 4.1 points of each other are inside the method’s demonstrated accuracy and should be treated as tied.
Where each candidate wins and loses
Every Democrat carries unaffiliated voters and the 1st District. The race is decided in the 2nd.
The same result, broken out by the groups that decide a Maine race: unaffiliated voters (a third of the electorate), the two congressional districts — the 1st around Portland leans Democratic; the 2nd is Trump country — and gender. The stronger candidates lose Collins by less in the 2nd District. Notably, Jared Golden — the sitting 2nd District congressman — posts the best showing there, as you would expect from real name recognition on home turf.
| Democrat | Statewide | Unaffiliated | ME‑01 | ME‑02 | Men | Women |
|---|---|---|---|---|---|---|
| Heather Cox Richardson | D+7 | D+19 | D+13 | C+2 | C+4 | D+9 |
| Shenna Bellows | D+5 | D+16 | D+11 | C+3 | C+6 | D+8 |
| Jared Golden | D+2 | D+13 | D+6 | C+5 | C+4 | D+1 |
| Troy Jackson | D+2 | D+12 | D+7 | C+6 | C+8 | D+4 |
| Patrick Dempsey | D+1 | D+13 | D+6 | C+6 | C+7 | D+3 |
| Nirav Shah | D+1 | D+13 | D+6 | C+7 | C+9 | D+3 |
| Dan Kleban | C+1 | D+10 | D+4 | C+9 | C+10 | D+1 |
| Jordan Wood | C+2 | D+8 | D+3 | C+10 | C+11 | Even |
| Valli Geiger | C+2 | D+8 | D+3 | C+10 | C+12 | Even |
D = the Democrat ahead; C = Collins ahead. All figures carry the same reality‑check adjustment as the headline numbers.
Why you can trust it
Five elections we already know the answer to
A poll is only as good as its track record, so before we asked about 2026 we made the same voter panel vote in five past Maine elections that have already happened and been officially counted. Because we know exactly how those turned out, we could grade the method against reality. It came close on all five — and it correctly captured the thing that makes Maine unusual: a large group of Democrats and independents who vote for Susan Collins even while backing Democrats elsewhere on the ballot. A method that missed that would badly misread this race.
| Certified race | Our method’s reading | Actual certified result |
|---|---|---|
| 2024 President (Harris / Trump) | Harris +5 | Harris +7 |
| 2020 President (Biden / Trump) | Biden +15 | Biden +9 |
| 2020 U.S. Senate (Collins / Gideon) | Collins +1 | Collins +9 |
| 2022 Governor (Mills / LePage) | Mills +20 | Mills +13 |
| 2018 Governor (Mills / Moody) | Mills +15 | Mills +8 |
The misses on this table are where the margin of error comes from: it is measured from how far the method landed from these five certified results, not assumed from a formula.
How it works
Real voters, a minimal profile, and honest corrections
Who we asked. A representative sample of 4,597 real registered Maine voters, drawn from the national voter file and covering all 16 counties, weighted so its mix of parties, ages, men and women, and backgrounds matches Maine’s actual electorate, county by county — standard practice for every credible poll.
How we asked. Instead of forcing each voter to just pick a name, we asked how likely they would be to vote for the Democrat on a 0‑to‑10 scale — letting people land in the middle or cross party lines, the way real voters do. Each matchup was asked on its own, so question order could not sway the answers.
The model sees almost nothing. Each simulated voter is a real Maine registration built from just a few facts: age, gender, race or ethnicity, and party registration. No names, addresses, or phone numbers are ever shown to the model, so no individual can be identified from an answer. County and district are used only to balance the sample and break out results. We deliberately tested a richer profile — modeled political lean, past turnout, income, education — and on the five certified elections it made the results slightly less accurate, so we kept the simpler, more transparent one. We let the track record decide.
The reality‑check correction. The one 2026 race with real public polling is Collins vs. Platner, which the polls show about tied. Our method guessed that same race and had the Democrat ahead by about 4 points — so we take about 4 points off every Democrat in these results to line up with that real‑world benchmark. A separate, independent estimate of the correction, built from the past elections alone, lands in the same place.
What to keep in mind. The method sees each candidate through a short description, so it treats all nine as fully established. The already‑known figures — a sitting congressman, a statewide official — sit close to where they really stand today. The lesser‑known names would start further back in a real 100‑day campaign, so their numbers assume the party gets behind them.
Run it on your race
This poll took no phone banks and no field time
The same method called 9 of 10 winners in the 2026 New York and Maryland primaries before a vote was cast. Any state, any matchup, any message — simulated against real registered voters and validated against certified results before you see a number.
Run date July 8, 2026 · simulated panel of 4,597 registered Maine voters drawn from the national voter file, all 16 counties · margin of error about 4.1 points, measured from the method’s misses on five certified past Maine elections · all figures carry a reality‑check adjustment anchored to current Collins–Platner public polling · prepared by Civly.