How does ranked-choice voting work?

The single-winner version used in U.S. elections is called instant-runoff voting. It runs a series of runoffs using the rankings voters already gave, so nobody has to come back for a second election.

Last reviewed · Nonpartisan explainer · Sources linked inline

Filling out the ballot

A ranked ballot is a grid: candidates down the side, rankings (1st, 2nd, 3rd...) across the top. You fill in one bubble per candidate and one per column. Your first choice is the one you want most; your second is who you'd want if your first can't win; and so on.

  • You don't have to rank everyone in most U.S. places. Ranking only one candidate is allowed; it just means your ballot stops counting if that candidate is eliminated. (Australia's federal lower house is the big exception: voters must number every box. AEC)
  • Ranking more candidates never hurts your first choice in the count itself. Your second choice is only looked at after your first choice is eliminated.
  • Giving two candidates the same rank is an "overvote." Most systems skip that ranking or stop counting the ballot at that point; the exact rule depends on local law.

The counting rules

  1. Count first choices. Every ballot counts for its top-ranked candidate.
  2. Look for a majority. If someone has more than half of the ballots still counting, they win. Done.
  3. Eliminate last place. Otherwise, the candidate with the fewest votes is removed. (Some places remove several hopeless candidates at once; see below.)
  4. Transfer. Each ballot that was counting for the eliminated candidate moves to its next choice who is still in the race. If none of its remaining choices are in the race, it is exhausted and sits out the rest of the count.
  5. Repeat steps 2 to 4 until someone has a majority.

Why "of the ballots still counting" matters

Exhausted ballots drop out of the total, so the majority line can move down from round to round. A winner can have a majority of the final round without having a majority of every ballot cast. That's one of the main points critics raise, covered in pros and cons.

A full worked example

One hundred voters, four candidates. These are fictional ballots; the count below is computed by the same code that runs our counter tool.

Four-candidate example: the ballots
Voters1st choice2nd choice3rd choice
34RiveraOkaforNguyen
8RiveraNguyen—
28NguyenOkaforRivera
18OkaforNguyenRivera
7OkaforRivera—
5PatelOkafor—
100ballots in total
Four-candidate example: the count, round by round
CandidateRound 1Round 2Round 3
Rivera 42 42.0% 42 42.0% 42 42.0%
Nguyen 28 28.0% 28 28.0% out
OkaforWinner 25 25.0% 30 30.0% 58 58.0%
Patel 5 5.0% outout
Exhausted (no longer counting)000
Ballots still counting100100100
  1. Round 1. 100 ballots are counting, so a majority is 51. Rivera leads with 42, short of a majority. Patel has the fewest votes and is eliminated. Those ballots move to their next choice: 5 to Okafor.
  2. Round 2. 100 ballots are counting, so a majority is 51. Rivera leads with 42, short of a majority. Nguyen has the fewest votes and is eliminated. Those ballots move to their next choice: 28 to Okafor.
  3. Round 3. 100 ballots are counting, so a majority is 51. Okafor has 58 (58.0%) and wins.

What the example shows

  • The first-round leader can lose. Rivera led the first two rounds but never reached a majority, because most voters who preferred someone else ranked Rivera low or not at all. Okafor started in third place and won with 58 of 100.
  • Every ballot counted in every round in this example, because every voter ranked at least two candidates and their next choice was still in the race.
  • Order of elimination matters. Patel went out first, and Patel's 5 voters lifted Okafor from 25 to 30, just past Nguyen's 28. Without them, Okafor would have been eliminated before Nguyen, and Okafor's voters would have decided between Rivera and Nguyen instead. Critics point to this sensitivity; supporters say it reflects how voters actually ranked.

Real elections where the first-round leader lost include the Maine 2nd District in 2018 and the Alaska House special election in 2022.

How real elections run the count

  • Batch elimination. When several low candidates together have fewer votes than the next-lowest candidate, they cannot catch up, so many places eliminate them in one round. New York City's 2025 Democratic mayoral primary dropped nine candidates in a single round (NYC Board of Elections). It doesn't change the winner; it only shortens the count.
  • Counting happens centrally. Because a ballot's next choice depends on who has been eliminated statewide or citywide, ballots (or their digital records) have to be gathered before ranked rounds can run. That is why final results can take longer. Alaska's 2022 special election was held August 16 and its ranked rounds were run August 31 (Ballotpedia).
  • First-choice results come first. Most jurisdictions release first-choice totals on election night, then run the rounds once all ballots, including mail ballots, are in.
  • Ties are broken by rules written in local law: often by earlier-round totals, and finally by lot. See exhausted ballots and ties.

Versions of ranked-choice voting

  • Instant-runoff voting (IRV) — one winner, described on this page. Used for Australia's House of Representatives and most U.S. ranked-choice elections.
  • Single transferable vote (STV) — several winners from one district, using a quota and surplus transfers. Used in Ireland, Malta, the Australian Senate and Cambridge, Massachusetts. How STV works.
  • Top-four or top-five primaries plus ranked general — Alaska's system pairs an open primary (the top four advance) with an instant-runoff general election.
  • Contingent and supplementary vote — older shortcuts that keep only the top two after the first count. London used the supplementary vote for mayor from 2000 until a 2022 law switched it to first-past-the-post (legislation.gov.uk).

Other ways of using ranked ballots, like the Borda count and Condorcet methods, count the same rankings differently and can pick a different winner. See them compared on the same ballots.

Common misunderstandings

  • "Some people get more than one vote." No. Each ballot counts for one candidate in each round. Courts have rejected "one person, one vote" challenges, including the Ninth Circuit in Dudum v. Arntz (2011) and a federal court in Maine in Baber v. Dunlap (2018) (Congressional Research Service).
  • "My second choice can hurt my first choice." In the count, your second choice is only read after your first choice is out. (There are rarer strategic effects, explained in voting methods compared.)
  • "It always picks the most broadly liked candidate." Not always. It can eliminate a candidate who would beat every rival head to head. Burlington, Vermont's 2009 mayoral race is the textbook case.