Ranking When It Counts

Ranking When It Counts

Katherine Gehl, Institute for Political Innovation

IPI Analytical Briefing · Updated November 13, 2026

What Alaska’s 2026 ballots show about voters who rank only one candidate

Prepared by Katherine Gehl (with Claude research and drafting assistance)

Internal analysis · Updated November 13, 2026 · Based on the Alaska Division of Elections cast vote record released November 12, 2026 (250,434 ballots, about 88% of the expected vote)

The bottom line

Critics will point out that about half of Alaska voters marked only one candidate in the big statewide races, and say this proves voters don’t want or understand instant runoff. The ballots say the opposite. Alaskans used the ranking option exactly when it could matter, and skipped it when it couldn’t.

  • Voters ranked when ranking could help them. In uncontested legislative races (one name on the ballot plus a write-in line), 1% ranked both the candidate and a write-in. With two candidates, 18% ranked; with three or more, 44%. Among voters whose first choice was eliminated in a statewide race, 58–66% had ranked a backup.
  • Two-thirds of all voters ranked somewhere. 66% of ballots ranked two or more candidates in at least one race that had three or more candidates.
  • Voters for a finalist lost nothing by stopping at one. Every one of those ballots counted in the final round. A single ranking is a complete, fully counted vote whenever your candidate is in the top two.
  • Instant runoff gave a voice to votes that pick-one plurality throws away. 4.1% of Alaska’s Senate voters (10,095) chose a candidate who couldn’t win. 58% of them (5,831) still helped decide between the top two. In pick-one plurality states this year, 4–14% of Senate votes had no say at all between the top two.
  • Exhausted ballots are mostly a choice, not a mistake. Nearly 9 in 10 exhausted statewide ballots belonged to voters who deliberately marked only one candidate. Overvotes and skipped rankings stopped only 0.4% (152 of 36,226).
  • Both parties ranked, and so did repeal supporters. Republicans ranked heavily for governor and Democrats for U.S. House; in the Senate the two sides ranked equally. Voters who supported repealing the system ranked more than those who voted to keep it (72% vs. 62%).

1. Voters ranked when ranking could make a difference

If voters disliked or misunderstood ranking, they would rank at similar rates everywhere. Instead, ranking rises sharply with the number of real choices on the ballot:

Type of race

Races

Ballots cast

Ranked 2 or more candidates

Legislative, uncontested (1 name + write-in line)

8

45,513

1.3%

Legislative, 2 candidates

24

158,347

18.3%

Legislative, 3 or more candidates

18

142,811

44.4%

Statewide (Governor, U.S. Senate, U.S. House)

3

743,791

42.2%

“Ballots cast” counts each ballot once per race. Candidate counts exclude write-ins. Source: author’s tabulation of the Division of Elections cast vote record.

The same pattern shows up within each statewide race. Voters whose favorite was a long shot ranked far more often than voters whose favorite was a frontrunner. They knew a backup choice was their insurance.

Race

Ranked 2+: first choice made the final two

Ranked 2+: first choice was eliminated

Governor

41.6%

66.0%

U.S. Senate

44.9%

63.1%

U.S. House

31.2%

58.1%

Source: author’s tabulation of the cast vote record.

A voter whose candidate makes the final two gives up nothing by ranking just that candidate. Every such ballot counted in the final round. For that voter, a single ranking is not a failure to participate; it is a complete vote.

2. Almost no one lost their voice, and those who did mostly chose to

An “exhausted” ballot is one that doesn’t count in the final round because it ranked no finalist. In the statewide races, exhaustion was small and overwhelmingly deliberate:

Race

First choice eliminated

Of those, still counted in final round

Exhausted (share of all ballots)

Exhausted ballots that ranked only one candidate

U.S. Senate

4.1% (10,095)

57.8% (5,831)

1.7% (4,264)

87%

U.S. House

10.2% (25,394)

54.7% (13,886)

4.6% (11,508)

92%

Governor

21.3% (52,414)

61.0% (31,960)

8.3% (20,454)

87%

All three

11.8% (87,903)

58.8% (51,677)

4.9% (36,226)

89%

Ballots stopped by an overvote or two skipped rankings: 152 of 36,226 exhausted statewide ballots (0.4%). Source: author’s tabulation.

The governor’s race shows what a deliberate exhausted ballot looks like. 98% of its exhausted ballots came from voters whose first choice was Bronson (R) or Taylor (R) and who chose not to rank Wilson (R). That is a voter saying “not her either,” which is their right. It is not confusion.

3. Pick-one plurality silences more voters than instant runoff

Under pick-one plurality, a vote for a candidate who can’t win has no say at all in the real contest between the top two. These are effectively exhausted ballots, and there is no way to rescue them. Compare Alaska’s Senate race with Senate races elsewhere this year:

2026 U.S. Senate race

Votes for candidates outside the top two

Share of those that still helped choose between the top two

Votes with no say in the final choice

Alaska (instant runoff)

4.1%

57.8%

1.7%

Montana (pick-one plurality)

14.0%

0%

14.0%

Ohio special (pick-one plurality)

4.8%

0%

4.8%

Massachusetts (pick-one plurality)

4.5%

0%

4.5%

Colorado (pick-one plurality)

4.4%

0%

4.4%

Unofficial results as reported by Ballotpedia (Montana, Ohio, Massachusetts, Colorado); totals may change before certification.

Montana is the starkest case. Republican Kurt Alme won with 48.3%, short of a majority. Independent Seth Bodnar (37.7%) and Democrat Alani Bankhead (12.1%) together had 49.8%. Under instant runoff, Bankhead’s voters would have helped decide the winner. Under pick-one plurality, their 58,540 votes had no say in that choice.

4. Both parties ranked; each ranked more where it helped

There is no sign that one party’s voters “got it” and the other’s didn’t. Each side ranked more when it had several candidates competing for the same voters:

Race

Ranked 2+: voters whose first choice for governor was Kreiss-Tomkins (D)

Ranked 2+: voters whose first choice for governor was a Republican

U.S. Senate (one main candidate per party)

44.6%

47.3%

Governor (three Republicans)

25.4%

68.1%

U.S. House (Begich (R) vs. Hill (N), plus Hafner (D))

42.9%

25.4%

Party side is based on each voter’s first choice for governor. Source: author’s tabulation.

Where the two parties were in the same position, as in the Senate race, they ranked at nearly the same rate. Republicans ranked heavily for governor because three Republicans were splitting their vote; the campaigns urged “rank 1, 2, 3.” Democratic-leaning voters ranked heavily for U.S. House because their candidate, Hafner (D), was running behind Hill (N). That is voters using the tool strategically and well.

5. Even voters who wanted to repeal the system used it

If ranking were a burden voters resented, the people voting to repeal it would be the least likely to rank. They were the most likely:

Ballot Measure 2 vote

Ranked 2+ in at least one race with 3+ candidates

Ranked 2+ in the U.S. Senate race

Yes (repeal)

72.0%

45.4%

No (keep the system)

61.6%

46.8%

All ballots

65.7%

45.7%

Source: author’s tabulation of the cast vote record.

Seven in ten repeal supporters ranked two or more candidates somewhere on their ballot. Whatever they thought of the system in principle, they used it in practice.

6. Legislative races: rankings decided outcomes

Legislative races are the hardest test, because voters know less about the candidates. Even so, in the 18 legislative races with three or more candidates, 44% of voters ranked. In two of them, the voters who ranked changed who leads, and two more are within 43 votes:

District

Final round (partial count)

Round 1 leader

Margin

Ranked 2+

Exhausted

Senate C

Stutes (R) 6,261 vs. Smith (R) 4,894

Stutes (R)

1,367

39.3%

6.5%

Senate E

Giessel (R) 8,773 vs. Stoddard (R) 4,062

Giessel (R)

4,711

36.0%

7.0%

Senate M

Berry (R) 6,282 vs. Tilton (R) 5,817

Tilton (R) → Berry (R) leads

465

50.1%

24.0%

Senate O

Rauscher (R) 7,198 vs. Sheldon (R) 5,633

Rauscher (R)

1,565

50.2%

21.9%

Senate S*

Edgmon (U) 2,239 vs. Deacon (R) 1,500

Edgmon (U)

739

43.9%

21.0%

House 2

Himschoot (N) 3,019 vs. Skaflestad (R) 1,888

Himschoot (N)

1,131

24.2%

3.4%

House 8

Elam (R) 6,368 vs. Savage (N) 1,923

Elam (R)

4,445

24.3%

3.1%

House 16

Hall (D) 4,345 vs. Vazquez (R) 2,813

Hall (D)

1,532

27.7%

2.2%

House 23

Allard (R) 3,647 vs. Arlington (D) 3,006

Allard (R)

641

61.2%

11.9%

House 25

Johnson (R) 3,559 vs. Bowles (R) 2,883

Johnson (R)

676

47.7%

23.2%

House 28

Wright (R) 2,860 vs. Moore (R) 2,817

Wright (R)

43

51.1%

19.6%

House 29

Nelson (R) 3,132 vs. Menard (R) 3,127

Nelson (R)

5

58.3%

26.0%

House 30

Massie (R) 3,183 vs. McCabe (R) 3,178

McCabe (R) → Massie (R) leads

5

50.5%

22.0%

House 31

Dibert (D) 3,332 vs. Moore (R) 2,019

Dibert (D)

1,313

32.0%

2.2%

House 33

Prax (R) 3,860 vs. McFarland (U) 2,139

Prax (R)

1,721

54.7%

9.4%

House 37*

Trefon (D) 876 vs. Smith (N) 650

Trefon (D)

226

38.6%

23.7%

House 38*

Jimmie (D) 1,230 vs. McCormick (D) 1,071

Jimmie (D)

159

53.9%

16.8%

House 39*

Foster (D) 1,479 vs. Ivanoff (N) 870

Foster (D)

609

48.0%

8.5%

* Mostly hand-counted rural districts; the cast vote record includes only part of their ballots. Party labels: R = Republican, D = Democrat, N = Nonpartisan, U = Undeclared. Results are partial; leads can change.

The higher exhaustion in some districts has a simple explanation. In the six districts with the highest exhaustion (Senate M and O; House 25, 28, 29 and 30), both finalists were Republicans. 89–96% of the exhausted ballots there came from voters who chose Kreiss-Tomkins (D) for governor, and most marked only their own candidate. They declined to choose between two Republicans. Those six races account for 71% of all exhausted ballots in multi-candidate legislative races.

And the Democratic voters who did rank in those districts are deciding the outcome. In House District 30, the Republican who trailed in round 1 moved ahead, by 5 votes, only after the Democrat's voters' backup choices were counted. In House Districts 28 and 29, those backup choices cut the leader's margin to 43 and 5 votes. In Senate M, the third Republican's voters flipped the lead. That is exactly the cross-party accountability the system is designed to create.

Anticipating the counterarguments

  • “Half of voters marked only one candidate in the big races.” Most of them were voting for a finalist, and every one of those votes counted in the final round. Ranking more would not have changed their influence.
  • “Exhausted ballots disenfranchise voters.” Statewide, 4.9% of ballots exhausted, and 89% of those voters deliberately marked one name. Under pick-one plurality, every vote for an also-ran is silenced in the final choice, with no option to rescue it.
  • “Voters are confused.” Overvotes and skipped rankings stopped only 0.4% of exhausted ballots, and voters’ ranking behavior tracked exactly where ranking could help them.
  • “Only one side benefits.” Both parties’ voters ranked heavily where it helped their side, and repeal supporters ranked at a higher rate than anyone.

About the data

  • Source: the Division of Elections’ cast vote record released November 12, 2026, covering 250,434 ballots (about 88% of the expected vote). It excludes about 6,000 counted ballots, mostly hand-counted in rural precincts, that had not yet been scanned. This version replaces earlier versions based on the November 4 (204,530 ballots) and November 11 (244,705 ballots) cast vote records.
  • Partial counts can change the order of elimination. In Senate District S, the November 4 file put Robb (N) in the final two; with more ballots counted, Deacon (R) replaced him. The final cast vote record is due November 18.
  • Tabulation follows Alaska’s rules: a ballot stops counting at an overvote or two consecutive skipped rankings; duplicate rankings of a candidate are skipped. Candidates were eliminated one at a time, lowest first, with all rounds run to the final two.
  • “Ranked 2 or more” counts distinct candidates ranked, including write-ins. Party side is the voter’s first choice for governor.
  • Figures will shift as remaining ballots are counted; the patterns are consistent across every race and are unlikely to change.

Glossary of terms

Instant runoff: The counting method in Alaska’s Final Four general election. Voters rank candidates; the last-place candidate is eliminated and those ballots move to their next choice, round by round, until two remain.

Pick-one plurality: The status quo in most states: voters choose one candidate and the most votes wins, even without a majority.

Exhausted ballot: A ballot that ranks no candidate still in the race and so doesn’t count in the final round.

Final round: The last round of counting, between the two remaining candidates.

Cast vote record: The election office’s ballot-by-ballot record of every vote, with no voter identity.

Overvote: Marking more than one candidate at the same ranking; under Alaska law the ballot stops counting at that point.

Sources

Download Full Paper Here