ANALYTICAL BRIEFING | OCTOBER 21, 2026

The best poll in the world: who won, the winning margin, and what voters want

Copyright Katherine Gehl · October 21, 2026

Summary

In an instant runoff election, most election offices stop counting the moment a candidate passes 50%. That tells us who won. It doesn’t tell us the winning margin, and it throws away most of what voters told us.

The winner is only the beginning. Every ballot carries a voter’s first, second, third and later choices. Counted all the way through, those rankings amount to an incredibly rich conversation between voters and the people who want to represent them. They show whether the winner has a mandate, where each candidate’s support really comes from, which issues resonated and which didn’t, and how voters weigh trade-offs.

Pollsters spend enormous effort trying to learn what voters want, and elected officials pay them to do it, by surveying a sample of people they can reach by phone. A fully counted instant runoff captures the actual, ranked preferences of everyone who showed up to vote. It is, in effect, the best poll in the world, and today most of it is never read.

Alaska’s 2022 elections show what we’re missing. Gov. Mike Dunleavy was reported as barely clearing a majority, with 50.3% of first choices. Run to the final two, he beat Les Gara 59.7% to 40.3%. In 12 legislative races in which the winner led after the first round, the same full count turned winners who looked alike, at roughly 50% to 62% of first choices, into a wide spread of winning margins, from 6 points to 34. And the movement of ballots between candidates shows which coalitions actually formed.

Bottom line

1. Not two results, but many

The winner. In an instant runoff, the winner is whoever reaches 50% plus one of the ballots still in the count. Once someone crosses that line, the winner is settled.

The winning margin. The winning margin is a separate result. In an instant runoff the clearest measure is the final round, between the last two candidates, after every other candidate’s voters have been asked who they prefer between the finalists. It answers a question pick-one plurality elections never can: when the field narrows to two, how broad is the winner’s support? A candidate can be a winner with a mandate or a winner without a mandate, and only running every round to the final two tells you which.

Under pick-one plurality, the winner and the winning margin come from the same single count: most votes wins, and the gap between first and second is the margin. In an instant runoff they are genuinely different results. A candidate must reach a majority to win, and there is a big difference between scraping past 50% and finishing the final round far ahead.

What voters want. Dig into how ballots move from candidate to candidate, and triangulate that against what each candidate ran on, and the rankings reveal far more:

• Where each candidate’s support comes from, and how deep or soft it is.

• Which issues resonated. Where did the voters who backed a candidate running on debt and deficits send their second and third choices: to a fiscal hawk from another party, or to a same-party candidate with different priorities?

• Which coalitions actually exist. Clusters of candidates whose supporters rank one another reveal issue coalitions that party labels hide.

• How voters weigh trade-offs, and how many found every finalist acceptable, or none.

This is a real back-and-forth between voters and the people who want to represent them, and it is valuable to everyone: to elected officials, who learn where their support comes from and what resonated; to candidates and parties; to journalists and researchers; and to the national conversation about what Americans actually want.

2. The Dunleavy case

In Alaska’s 2022 general election for governor, Mike Dunleavy won 50.3% of first choices in the certified count. Because he had a majority, the Division of Elections’ official tabulation stopped after round one, and the public story was that he barely won.

Using the state’s public cast vote record, the Ranked Choice Voting Resource Center ran every round.¹

• Run to the final two, Dunleavy beats Gara 59.7% to 40.3%, a margin of more than 48,000 votes.

• The rankings tell us who Alaskans were choosing between. Of Republican Charlie Pierce’s 12,164 ballots, 7,149 moved to Dunleavy and 648 to Gara. Independent Bill Walker’s 56,144 ballots broke for Gara by more than four to one (35,059 to 8,072), and 13,013 ranked neither finalist. That is a portrait of two distinct coalitions, and of a sizable bloc of Walker voters who wanted neither major-party finalist.

• Very few ballots dropped out. Only 17,103 of 266,693 ballots (6.4%) were inactive by the final round.

• Same winner, very different result. If the final round had still been near 50–50, Dunleavy would have been a winner without a mandate. Instead, the full count shows a decisive mandate.

3. Beyond the governor: 12 legislative races

The Ranked Choice Voting Resource Center ran the same full count for 12 Alaska legislative races from 2022 in which the winner led after the first round. In the first round, nearly every winner looked alike, with roughly 50% to 62% of first choices. The final two tell a much more varied story.

Round-1 shares and final-two results are from the cast vote record, run to two candidates. In Senate J, the winner’s 49.9% rose above 50% once write-ins were removed in round 2.

4. How the full count was done

No special access was needed. The Alaska Division of Elections publishes its cast vote record, the anonymized ballot-by-ballot record of every ranking, after each election. It contains no individual voter information: no names, addresses or voter ID numbers. The full count was run with RCTab, free, open-source tabulation software hosted by the Ranked Choice Voting Resource Center, which has been federally tested and is certified for official use in New York and Utah. The only setting that differed from Alaska’s official tabulation was the instruction to keep counting until two candidates remain.

Anyone can reproduce these results from public data. That is exactly why election offices should release the cast vote record, and why they should run, and report, all the rounds themselves.

5. Best practice

6. The best poll in the world

Campaigns and officeholders spend heavily on polls to learn what voters think. A fully counted instant runoff answers many of the same questions, more completely and at no extra cost:

Two cautions keep the analysis honest. Ballots rank candidates, not issues, so reading issue signals requires triangulating the rankings against what each candidate ran on. And the data is anonymous: the cast vote record contains no names, addresses or voter ID numbers, only how ballots were marked.

Pick-one plurality elections can’t produce any of this. Many votes are lesser-of-two-evils choices, and most races were effectively decided in the primary, so the general election reveals little about what voters truly prefer. Instant runoff, run all the way through, turns every election into a detailed, honest reading of the electorate. IPI will explore this voter information result in depth in a future briefing.

7. Likely objections and questions

Sources

• Alaska Division of Elections, 2022 general election results and cast vote record

• Chris Hughes, Ranked Choice Voting Resource Center: full-round RCTab tabulations of the 2022 Alaska governor’s race and 12 legislative races (“2022 Final Two Results,” June 12–28, 2024), prepared at IPI’s request

• Ranked Choice Voting Resource Center, RCTab

• PBS NewsHour/AP, Dunleavy becomes first Alaska governor reelected since 1998 (Nov. 23, 2022)

• Unite America, 87% of U.S. House elections already determined in primaries (Oct. 8, 2024)

• FairVote, When should we expect results in Alaska? (tabulation time) and Alaska cast vote record released (Sept. 2022)

Glossary of terms

Terms as used in this briefing. For the full set, see IPI’s Political Innovation—A Guide to Terms.

Cast vote record (CVR). The electronic, ballot-by-ballot record of the choices marked on every ballot, produced by the tabulating equipment as ballots are scanned. In an instant runoff race, each record shows a ballot’s first, second, third and later rankings. A cast vote record contains no individual voter information: no names, addresses or voter ID numbers, and it is not linked to the voter rolls. With it, anyone can run the instant-runoff rounds and reach the same answer the state will.

Exhausted (inactive) ballot. In an instant runoff, a ballot that stops counting in later rounds because every candidate it ranked has been eliminated (or it was spoiled, for example by an overvote). Alaska law calls these inactive ballots. Under pick-one plurality, a vote for a losing candidate also has no effect on the outcome; it simply isn’t counted as “exhausted.”

Final Five Voting (Final Five Elections). A political innovation designed by Katherine Gehl that replaces party primaries and pick-one plurality general elections with two changes: (1) a single, open top-five primary in which all candidates appear on one ballot and the five with the most votes advance, and (2) an instant runoff general election, in which voters rank the finalists and the winner must earn a majority.

Final Four (Alaska). Alaska’s version of the system, adopted by ballot measure in 2020 and first used in 2022: a top-four open primary followed by an instant runoff general election.

Winning margin. The margin by which the winner wins, measured in the final round between the last two candidates. It is a separate result from who won, and it shows whether the winner has a mandate. The winning margin is known only after every round is run to the final two.

Instant runoff. The ranked general election used inside a Final Five (or Alaska’s Final Four) system. Voters rank candidates; if no one has a majority of first choices, the last-place candidate is eliminated and those ballots move to their next choice, round by round, until a candidate has a majority. IPI uses “instant runoff” for this use and “ranked choice voting” for standalone ranked elections.

Majority winner. A candidate who wins more than half of the votes. Some states reach a majority winner through instant runoffs, counted from a single ballot (Alaska for state and federal general elections; Maine for party primaries and federal general elections). Others reach it through physical runoffs, separate elections held weeks later (Georgia, for example).

Mandate. The strength of a winner’s support. Under instant runoff, the clearest measure is the winning margin between the last two candidates. A candidate can win with a mandate or without one.

Physical (separate) runoff. A second, separate election held weeks later when no candidate wins a majority or a set threshold. Georgia holds its general election runoff four weeks after Election Day; its 2022 U.S. Senate race was decided in a Dec. 6 runoff. An instant runoff produces a majority winner without a second election.

Pick-one plurality. Today’s most common way of electing officials. The term describes both halves of the system. “Pick one”: voters choose only one candidate and cannot rank anyone. “Plurality”: the winner is the candidate with the most votes, not necessarily a majority. IPI uses the full phrase, pick-one plurality, so both features are always in view.

Plurality winner. A candidate who wins the most votes, even if that is less than half. Winning by plurality is what most U.S. general elections have traditionally required. In a crowded field, a plurality winner can be opposed by most voters.

RCTab. Free, open-source software, hosted by the Ranked Choice Voting Resource Center, that tabulates instant runoff and ranked choice results from a cast vote record. It has been federally tested and is certified for official use in New York and Utah. Anyone can download it and run the rounds.

Round (elimination round). One step of an instant runoff count. In each round the last-place candidate is eliminated and those ballots count for their next-ranked choice still in the race.

Run all the rounds. The practice of continuing an instant runoff count until only two candidates remain, even after a candidate has passed 50%. It doesn’t change who won; it reveals the winning margin.

Tabulated result. In instant runoff races, the result produced when the state runs enough rounds to determine a 50%-plus-one winner. A state may choose to tabulate once, daily or on another schedule. IPI’s recommended practice is for the state to tabulate once, when every ballot is in, while releasing the cast vote record along the way so decision desks can run the rounds and call races.

Voter information result. The layer of results produced by reading how voters distribute their second, third and later rankings in a Final Five or Final Four election, triangulated against what candidates ran on. It shows where each candidate’s support comes from, which issues resonated, which coalitions exist and how voters weigh trade-offs, across the entire electorate that voted. IPI calls a fully counted instant runoff “the best poll in the world.”

Winner. Whoever reaches 50% plus one of the ballots still in the count. The winner can often be called before the winning margin is known.

About the Institute for Political Innovation

The Institute for Political Innovation (IPI) is a cross-partisan, not-for-profit organization founded by Katherine Gehl in 2020. IPI provides research-based ideas, scholarship and strategy to update American politics, applying the lens of industry competition to the politics industry and its rules. IPI Analytical Briefings share research and analysis on how elections machinery and legislative machinery shape the results that state and federal governments ultimately deliver for their customers—the citizens.

political-innovation.org

Notes

1. All figures in this briefing come from full-round tabulations run on June 12, 2024 by Chris Hughes, Senior Director of Policy and General Counsel at the Ranked Choice Voting Resource Center, at IPI’s request. He used RCTab with the rule “continue until two candidates remain,” applied to the Alaska Division of Elections’ public cast vote record export for the Nov. 8, 2022 general election. Ballot-level counts differ slightly from certified first-choice totals (for example, Dunleavy’s certified first-choice count was 132,632, or 50.29%), which is common between cast vote records and certified canvasses.

   ↩