Executive summary

Ranked choice voting (RCV) uses a round-by-round count of ballots to determine winners in both single-winner and proportional (multi-winner) RCV elections. In an RCV tabulation, the candidates with the least support are eliminated, ensuring that the candidate or candidates who get the most votes win. Ranked choice voting reduces wasted votes and encourages people to vote sincerely, rather than strategically.

In any election, however, one thing remains the same: the public has a strong interest in learning the results. Because RCV is a less common method to identify who won, the process for releasing results in RCV races can be especially important to ensure understanding, to convey results in a way that gives people the information they need, and to ensure the outcome is trusted and understood.

FairVote and the Ranked Choice Voting Resource Center have analyzed results from hundreds of RCV contests. Based on that experience, we recommend the following tried and true tips:

1. Release a preliminary round-by-round tally on election night.
2. Continue to release preliminary tallies as more votes are counted.
3. Conduct vote total checks with each release of preliminary results.
4. Include ballots with write-in candidates as a single batch in the first round.
5. Publish the full ballot record so that anyone can verify the result.
6. Make use of existing tools for visualizing RCV results.
7. Clearly communicate expectations, timelines, and results.

Following these best practices – to the extent permitted by state law – can help instill public confidence in the electoral process and its outcome.¹

Recommendation 1: Release a preliminary RCV tally on election night

With or without RCV, counting ballots takes time, especially as voting options have expanded beyond casting a ballot at a local precinct on Election Day. However, Americans have become accustomed to getting election results quickly, with candidates holding rallies on election night, and with journalists eager to be the first to announce the winner. Final results take time, but regularly releasing preliminary results has become a normal practice, helping campaigns and the media get an accurate sense of whether they can safely project a winner or if a contest will be too close to call until more results come in. Preliminary results convey the same benefits in RCV races as they do in single-choice races. However, early results can shape opinions about the legitimacy of final results, so they must be as accurate as possible.

When using RCV, the best way to ensure early results accurately convey the state of the race is to release a round-by-round RCV tally, rather than merely first-choice results. Compare media attention around two similar RCV contests: the June 2018 mayoral special election in San Francisco, where a preliminary RCV tally was released on election night, and the June 2021 mayoral primary election in New York City, where only first choices were released prior to final results.

In the San Francisco contest, London Breed led her closest competitor, Mark Leno, by a 10-point margin in first choices on election night, after a count of about 30% of the total ballots cast.² Because San Francisco released a preliminary RCV count of those ballots on election night, however, it was clear that Breed’s wide margin in the first round would shrink in the RCV count – in the final round, Leno had narrowed the gap to less than 5 points. 

As more ballots were counted, San Francisco continued to update its preliminary reports, releasing two reports on election night, and an additional report each day after that. Each report reinforced that the race would be very close. The last preliminary report on election night maintained Breed’s 10-point margin in first-choices, but showed Leno winning in the final round by less than a single point, with just over 60% of votes counted. Contemporaneous media reports accurately conveyed that the race was close and that no candidate could fairly be called a frontrunner until all the ballots were counted.³ In the final report, Breed regained her lead, and won with 50.55% of the final-round vote.⁴

In New York City’s contest, the fundamentals of the race were quite similar. On election night, first-choice results showed Eric Adams beating his closest competitor in the first round, Maya Wiley, by a margin over 9 points.⁵ However, New York City did not release a preliminary RCV count on election night, so no one knew whether that lead would last through the round-by-round count.

Although Adams had earned only 31.7% of first choices, media sources quickly and repeatedly referred to Adams as the frontrunner, and referenced his “commanding lead.”⁶  And while some analysts, citing FairVote data, noted that in prior RCV contests, margins had shifted by more than 10 points as candidates were eliminated,⁷ this was not the dominant media narrative. 

When the first preliminary RCV tally was released a week later, it showed Adams winning in the final round by a margin of less than one point over Kathryn Garcia, who placed third in the initial round, but passed Wiley in the RCV count.⁸ Had Adams lost in the final round, and lost to the candidate who placed third in first choices, the surprise result could have caused confusion and undermined confidence in the legitimacy of the outcome.⁹ Had New York City released a preliminary RCV count on election night, the media would have known that the contest was in fact a very close race between Adams and Garcia, and reporting would have reflected that.

Quickly releasing preliminary round-by-round results offers greater transparency and facilitates broader understanding of, and acclimation to, ranked choice voting. Preliminary results should always be clearly labeled as preliminary, indicate that the order of elimination may change as ballots are counted, and if possible, include the percentage of ballots counted to date.

To the extent permitted by law, preprocessing of mail-in ballots can be especially helpful to allow for a timely and robust preliminary count.

Recommendation 2: Continue to release preliminary tallies as more votes are counted

Voters have come to expect preliminary results that are updated as ballots are counted in non-RCV races. The same policy should be followed for RCV races as well. 

In Boulder, Colorado’s first use of RCV in 2023, the city not only released three sets of preliminary results on election night, but also continued to update its preliminary results on a daily basis over the next two days.¹⁰ This helped to normalize the RCV counting process, giving the media new material to report on without the need for speculation. It also helped to dispel the myth that RCV delays election results. The public could readily see that ballots were continuing to arrive and were being counted according to Boulder’s policies, while the round-by-round RCV count required only the push of a button.

New results do not necessarily need to be daily, so long as the update schedule is published ahead of time, and updates are regular and frequent enough to demonstrate transparency in the counting process. As with preliminary round-by-round tallies on election night, new tallies should always be clearly labeled as preliminary, and if possible, include the percentage of ballots counted to date.

Recommendation 3: Conduct vote total checks with each release of preliminary results

To minimize the risk of posting erroneous data, results should always be double-checked against accurate vote totals from a trusted source, such as an election management system or early canvass reports, before publishing preliminary or final results.

When Maine conducted its July 2020 primary elections, two such errors occurred: Staff at one point uploaded the wrong file, and a memory device malfunctioned and only uploaded some of its ballot data.¹¹ Similarly, when New York City used RCV in June 2021, the city mistakenly included over 100,000 sample test ballots in its first batch of round-by-round results.¹² In both cases, election administrators responded appropriately: The discrepancy was identified and corrected quickly and transparently. Nonetheless, these errors could have been avoided through best practices. (Of course, uploading the incorrect file or a temporary tech malfunction have nothing to do with RCV; these types of human or temporary tech errors can also occur in choose-one elections.)

Because posting erroneous results can undermine public confidence in the counting process, RCV tallies should always be spot-checked against expected vote totals to ensure the data matches.

Recommendation 4: Count ballots with write-in candidates as a single batch in the first round

All ballots should be counted in the first round for their highest-ranked candidate, including ballots with a write-in ranked first.¹³ Ballots with first-choice write-ins can be counted in one batch in the first round. If the first-round total for write-in candidates is smaller than the totals for every other candidate, or if a jurisdiction has no filed write-in candidates, the write-in candidates should all be eliminated in the first round, with those ballots transferring to voters’ next ranked candidate.

Because first-round totals are used to determine candidate elimination, the best practice is to ensure all ballots with a first-choice write-in are counted in the first round.¹⁴

Recommendation 5: Publish the full ballot record so that anyone can verify the result

A cast vote record (CVR) is a digital anonymized file containing information on the ranking order on each ballot cast. It has become the norm for election officials in RCV jurisdictions to publish the CVR for external verification and analysis, creating maximum crowd-sourced transparency and election security. Maine, Alaska, New York City (NY), Berkeley (CA), Oakland (CA), San Leandro (CA), Albany (CA), Minneapolis (MN), Portland (ME), St. Louis Park (MN), Bloomington (MN), Minnetonka (MN), Boulder (CO), Burlington (VT), Easthampton (MA), and Cambridge (MA) all publish their CVRs once results are complete, and San Francisco (CA) publishes a CVR with each release of results, including preliminary results.

We recommend the release of a CVR as a best practice for all RCV contests. This practice ensures that election results are viewed by the public as transparent, credible, and legitimate.¹⁵ The final CVR should certainly be released along with certified results, but jurisdictions should follow San Francisco’s practice of releasing preliminary CVRs at regular intervals before certification as well to allow for independent verification, if feasible. Cast vote records should also be reviewed before release to ensure voter privacy is maintained. If jurisdictions have precincts with low turnout (10 or fewer votes cast), votes from those precincts should be combined to preserve voter privacy.

Recommendation 6: Make use of tools for visualizing RCV results

To have confidence in the result of an election, voters should understand not only who won, but how they won. In a single-choice election, voters expect to know the breakdown of votes. With RCV, voters should understand the round-by-round count. Because round-by-round counting of ranked ballots remains unfamiliar to many, it should be presented in a way that is accessible and transparent, and multiple ways of visualizing RCV results have evolved to accomplish that goal.

These visual RCV election displays have been built into some software provided by voting machine vendors; some results reporting vendors, such as Enhanced Voting, have RCV results display tools too. Additional development is required to provide wide access to RCV results display tools, however. Fortunately, independent software developers have created free tools that can import ballot records and visualize RCV results following best practices from usability testing.

For example, Boulder County, Colorado links from its official results page to RCVis.com, a website where RCV ballot data can be used to generate visualizations of results in a variety of formats.¹⁶ The two images below show the results for the Boulder mayoral election as visualized by RCVis.¹⁷ These results formats give voters the information they need to identify the winner at a glance and understand how that winner won. They follow some basic best practices, such as emphasizing the decisive final round of the tally while also showing how vote totals changed over the course of the round-by-round count. We suggest that jurisdictions embed these RCV results displays on their own websites.¹⁸

One element to consider when visualizing RCV results is whether to continue rounds of counting until two candidates remain, even if the ultimate winner earned a majority in an earlier round. This practice provides a consistent way of showing a winning candidate’s “mandate” versus their strongest opponent, but it also increases the number of inactive ballots in the final round. San Francisco adopted the practice of running down to two candidates beginning in 2015, and New York City requires it by law; Utah cities, on the other hand, stop the round-by-round counting process once someone achieves a majority of the active votes.¹⁹ The Ranked Choice Voting Resource Center believes that each approach has its own benefits, and FairVote recommends that counting continue until only two candidates remain.

Newer and clearer ways of presenting RCV results may become available over time, and jurisdictions should always use the best tools available to them. For now, we recommend jurisdictions rely on the approaches identified in the Center for Civic Design’s Best practices for ranked choice voting ballots and other materials report, and use the display tools available at RCVis.com.

Recommendation 7: Clearly communicate expectations, timelines, and results

Wherever there is ambiguity, there is room for spin and speculation. To increase clarity and public confidence, election officials should publish and publicize a clear plan well before Election Day, and then provide regular updates on that plan’s progress and any unexpected changes to it. That plan should include an expected timeline for releasing results and should clearly define the frequency and cadence for releasing unofficial round-by-round results on election night and at regular intervals thereafter. 

Most voters do not look for information about elections on official government websites, but by watching their preferred media sources and checking social media. Officials should therefore be proactive about explaining the timeline and process with the media before Election Day, posting status updates across all social media channels on and after Election Day, and sharing best practices for media displays of RCV results.²⁰ To this end, officials should also host a test run of the results reporting process for the press, stakeholders, and the public prior to the election. This gives everyone a chance to ask questions and learn about how results will be produced, while also testing the process, system, and technology.

References

1. For more information on administering RCV elections, we recommend two additional resources. For best practices in displaying results, see Best practices for Displaying Ranked Choice Voting Election Results, Center for Civic Design, (2022). For an overview of reporting practices in RCV jurisdictions, see Reporting the Results of Ranked-Choice Voting Elections: Successes and Pitfalls Across Forty-Six RCV Jurisdictions, Equal Democracy Project at Harvard Law School & Ranked Choice Voting Resource Center (Aug. 2022), https://www.rcvresources.org/blog-post/reporting-the-results-of-ranked-choice-voting-elections.

2. June 5, 2018 Election Results – Detailed Reports, San Francisco Department of Elections. 

3. See, e.g. San Francisco mayor’s race very tight under unusual system, ABC 10, Local (June 6, 2018).

4. June 5, 2018 Election Results – Detailed Reports, San Francisco Department of Elections.

5. See Aaron Blake, Eric Adams and the history of come-from-behind wins in ranked-choice races, Washington Post (June 23, 2021).

6. E.g. Joshua Chaffin, Eric Adams storms to lead in New York City mayoral race, Financial Times (June 23, 2021).

7. See Blake, supra n.9.

8. Official Ranked Choice Rounds, Board of Elections in the City of New York, 2021 Primary.

9. Eric Durkin, New York’s first full ranked-choice election changed campaigns — if not the results, Politico (August 24, 2021).

10. Ballot Processing Election Night & Beyond. Boulder County.

11. Secretary of State Matt Dunlap Announces Corrected Tabulation for Races Decided by Ranked-Choice Voting. Department of the Secretary of State of Maine. (August 3, 2020).

12. Karen Matthews, NYC mayor race: Test ballots wreak havoc, lead to 135K ballot ‘discrepancy’ in ranked-choice results, CNBC Politics (June 30, 2021).

13. Write-in processing policies vary widely in the U.S., whether jurisdictions are using RCV or not. Wherever possible, we suggest jurisdictions eliminate write-ins as a batch as discussed here. Jurisdictions that need to adjudicate all write-in data (because the contest has a declared write-in candidates or other local policies) may not be able to follow this best practice. In that instance we suggest clearly communicating with the public about how the write-in process may impact results timing. Ideally, more efficient write-in processing policies can be adopted in advance of RCV implementation as well.

14. Some ranked choice voting software includes an option for “first round suspension” but this does not count write-ins in the first round. If this option is selected, any ballots whose first choice is a write-in or a skipped ranking will not be counted in the first round, but will be counted in subsequent rounds. There is no jurisdiction in the U.S. whose tabulation rules call for the use of first round suspension.

15. See Steve Mistler, ‘It’s Just Math’ — Mainers Recreate Ranked-Choice Voting Results To Test The Process, Maine Public (December 2, 2018).

16. See 2023 Coordinated Election Official Results. Boulder County Elections. (December 6, 2023).

17. City of Boulder Mayoral Candidates.

18. RCVis provides iframe embedding capability.

19. Compare the results from New York City’s March 2021 11th council district special election, which continues until two candidates remain with Vineyard, Utah’s November 2021 city council election, which stops when the winner has a majority in the second round.

20. See How to display RCV election results. FairVote. (August 30, 2022.)

About the authors

This report was co-authored by FairVote and the Ranked Choice Voting Resource Center.