Instant Runoff Voting and Its Impact on Racial Minorities
Produced by The New America Foundation and FairVote, June 2008

Download this report as a PDF.

Executive summary:
Instant runoff voting (IRV) is an election method that determines the choice of a majority of voters in a single round of voting without the need to conduct a separate runoff election. As a majority voting method, IRV is ideal for single-winner offices such as governor, mayor and legislators representing a one-seat district.

In recent years, IRV has been implemented for local elections in several American cities, including San Francisco (CA), Cary (NC), Hendersonville (NC), Takoma Park (MD), and Burlington (VT). It has also been used for state elections to allow overseas voters to participate in runoff elections in such states as Louisiana, South Carolina, and Arkansas.[1]  In 2008 Pierce County (WA) began using IRV to elect the county executive and other county offices. IRV has been popular with voters, having won in 13 out of its last 14 ballot measures across the country, including landslide victories in 2006-2008 that averaged 69% of the vote in Oakland (CA), Minneapolis (MN), Santa Fee (NM) and Sarasota (FL).

Internationally, IRV is used to elect the Mayor of London, an extremely diverse and complex city with many racial and language minorities. IRV has been used for generations in national elections in Australia and Ireland. In 2007 it was used for highly praised elections in the complex, multiracial electorate of Papua New Guinea. The related ranked voting system of choice voting (also known as proportional representation) has been used in Cambridge (MA) since 1941 and was previously used with positive impact on representation of racial and ethnic minorities in such cites as New York, Cleveland and Cincinnati.

San Francisco is the largest and most diverse American city to hold an IRV election in recent years. The city’s population is 32% Asian American, 7% African American, 14% Latino and 44% white (2006 ACS estimate); a majority of residents aged 25 and older do not have a college degree. San Francisco has administered instant runoff voting in annual elections for numerous local offices since November of 2004. The Public Research Institute at San Francisco State University conducted exit polls during two IRV elections in this period, as did other groups like the Asian Law Caucus. In addition, FairVote conducted studies based on precinct analysis of ballots for the 2004 and 2005 elections. The polls of San Francisco voters’ opinions and the precinct analysis of turnout and use of rankings in San Francisco demonstrate that:
  • Voters of all races and ethnicities strongly preferred IRV over a two-round runoff system.
  • Voters of all races and ethnicities find IRV easy to use.
  • IRV increased turnout citywide by 2.7 times, and in the city’s six most racially and socio-economically diverse neighborhoods turnout quadrupled in the 2005 citywide election (the only race studied for impact on voter turnout).
  • An overwhelming majority of voters, including minority voters, reported understanding IRV.
This report concludes with a review of voter education efforts in two IRV elections, including San Francisco’s well-funded voter education and outreach campaign in 2004 and Cary, NC’s modestly-funded yet still effective voter education effort in 2007.

A. Voters preferred instant runoff voting

San Francisco’s exit poll results show that voters smoothly transitioned to IRV. The Public Research Institute at San Francisco State University (SFSU) conducted exit polls during the 2004 and 2005 elections that used IRV, including both district races for the board of supervisors and citywide races for various offices. According to the SFSU surveys, voters preferred IRV to the previous runoff system by a margin of over three to one. Overall, 55% of voters stated a preference for IRV, 28% stated no preference, and only 17% missed having the two-round runoff.[2]  In every single grouping of voters, as measured by age, race, language, income and education, voters preferred IRV to the old delayed runoff method.[3]  

In 2006, the Asian Law Caucus conducted an exit poll of voters for a Board of Supervisors race in District Four, a majority Asian district. Four different Asian candidates were running in this race, which would normally raise concerns about the possibility of Asian voters "splitting their vote" among too many Asian candidates. IRV allowed the Asian vote to coalesce on a winning Asian candidate. The ALC survey found that 66% of Asian respondents called IRV “helpful” compared to 57% of non-Asian respondents. And 82% of Asian respondents ranked two or three choices, while 84% of non-Asians ranked two or three choices.

Exit polls in other U.S. cities that have used IRV for local elections report similar findings. In Cary (NC), 68% of voters preferred IRV over their previous method; in Burlington (VT) 63%, Takoma Park (MD) 89% and Hendersonville (NC) 67%.[4]  

B. Voters report a high level of understanding

While an understanding of IRV's vote tabulation rules are not necessary for a voter to rank candidates and cast an effective ballot[5], exit polls typically have asked questions to voters about their level of understanding of IRV. In SFSU’s exit polls in 2004 and 2005, for example, 87% of San Franciscans polled said they understood the voting system “perfectly well” or “fairly well”[6].  Specifically, ethnic and minority voters reported a very high level of understanding of IRV:[7]

 Understand "Perfectly well" or "Fairly Well" 
 Hispanic/Latino 89.8%
 Asian/Pacific Islander
 African American/Black
 White 87.8%
 Other 85.3%

Exit polls in other American cities using IRV in their local elections also showed that voters reported high levels of understanding: 88% in Takoma Park, MD,[8]  95% in Cary, NC ,[9] 86% in Hendersonville, NC,[10]  and 86% in Burlington, VT.[11]  

C. IRV did not increase the rate of spoiled ballots

The San Francisco studies show that regardless of whether voters knew that they would be asked to rank candidates when they showed up to vote, it was equally easy for them to cast a ranked ballot. Nearly half of San Francisco voters in the city’s IRV election in 2005 did not know they were going to vote in an IRV election when they came to vote or received their absentee ballot, yet 99.6% of voters in that year’s most hotly contested IRV race cast valid votes – with all categories of voters handling IRV elections well whether they knew about it before voting or not.[12]

In IRV elections, if voters erroneously mark more than one first-choice candidate, it is invalidated as an “overvote” Overvotes are "spoiled" and indicate a lack of voter understanding of how to mark their ballot. If voters do not vote for any candidate in a given race, it is called an “undervote.” Undervotes are not "spoiled" ballots. They typically reflect a voter’s intent to skip voting in a race due to indifference to the outcome.

In the November 2004 election, San Francisco held IRV elections for seven seats on the Board of Supervisors, along with non-IRV state and federal elections. This election provided valuable data for a precinct analysis by FairVote that compared IRV races with non-IRV races. For the non-IRV presidential, federal, and state races, less than 1% of voters overvoted and 8% of voters undervoted, for a countable-ballot rate of 91.75%. In all the non-presidential races, the countable-ballot rate was 89.7%.[13]  

In contrast, in the seven IRV races for the Board of Supervisors, less than 1% of voters overvoted and 8% of voters undervoted, for a countable-ballot rate of 91.1% -- virtually identical to the non-IRV rate, and in fact higher than the countable-ballot rate in non-Presidential elections overall.

IRV elections have shown a very low overvote rate across the nation. In Burlington (VT), for example, 99.9% of ballots cast in a hotly contested five-candidate race for mayor in 2006 were valid. The undervote was lower in the mayor’s race than the other Burlington elections that year that did not use IRV, and the turnout was significantly higher than the preceding mayoral election without IRV. In Takoma Park (MD), there was only a single overvote in the city’s initial 2007 IRV election with three candidates seeking an open seat -- and there was only a single overvote among all the second choice rankings cast as well.

D. IRV increased voter turnout

FairVote’s precinct analysis of San Francisco’s first citywide election with IRV showed that using IRV rather than a two-round runoff election significantly increased voter turnout in the decisive election, especially among ethnic and racial minority groups. The FairVote precinct analysis compared the 2001 election for City Attorney that did not use IRV (as it had not been passed by voters yet) with a comparable 2005 election for City Attorney and Assessor-Recorder, using IRV. The city attorney race in 2005 was uncontested so a comparison was made between the city attorney race in 2001 and the Assessor-Recorder race in 2005. Both elections took place in odd years and both offices held a similar public profile. The runoff in the December 2001 election for City Attorney had a voter turnout of just 17% of registered voters (74,698 out of 453,961 registered voters).[14] In contrast, during the November 2005 IRV election, turnout was 54% of registered voters (229,714 out of 428,481 registered voters).  Based on this analysis, IRV boosted citywide voter turnout in the final, decisive race nearly three-fold (by 2.7 times).[15]

Significantly, the most dramatic rise in turnout was in neighborhoods where people of color make up a majority.[16]  Before IRV was introduced, San Francisco’s most diverse neighborhoods had the lowest turnout rates. After IRV was introduced, voter turnout in the city’s six most socio-economically diverse neighborhoods skyrocketed by over 300% (a four-fold increase in turn out): Western Addition (309.4%), Bayview / Hunter’s Point (351.6%), Mission (351.6%), Ingleside (324.6%), Excelsior (310.4%) and Visitation Valley (407.3%).[17]

Another FairVote study comparing San Francisco’s voter turnout in two presidential election years, 2000 and 2004, also showed a dramatic increase in participation resulting from the use of IRV. In the Board of Supervisor races in 2000, IRV was not used and voter turnout declined significantly between the November election and the December runoff. The December runoff turnout ranged from a low of 53% to a high of 64% of the November turnout, depending on the race. In IRV races for Board of Supervisor in the 2004 election, final round turnout ranged from a low of 75% to a high of 90% of the initial round of voters who cast ballots.

In raw numbers, winning candidates in the 2000 elections with December’s delayed runoffs received between 5,900 and 10,400 votes. Winning candidates in the IRV elections in 2004 received between 10,500 and 13,600, and were ranked in the top three on 12,200 to 16,900 ballots. That means that when comparing the IRV elections to the December runoff elections, many more voters were participating in the final round of the IRV election when the race was decided than were voting in the December runoff elections. And it means with IRV more voters were having a direct impact on the choice of their elected representative than with the December runoffs.

Two key factors account for this dramatic difference in voter turnout: (1) consolidating from two elections to one relieves voter fatigue, and enables voters, campaigns and mobilizing organizations to focus on a single, decisive election, and (2) in San Francisco’s case, holding the final election in November rather than December takes advantage of typically higher turnout in November elections when higher profile races like president and governor are on the ballot.

E. Racial minorities more likely to use rankings

With IRV, voters have the options of ranking up to three candidates. That is, they can rank three candidates, rank two candidates, or simply mark a single candidate (as their first and only choice). According to the San Francisco State University study of the 2005 IRV election, San Francisco’s racial and ethnic minority voters ranked the most candidates in the citywide election. They overwhelmingly chose to rank three candidates -- more so than white voters:

 Ranked Three 
Hispanic/Latino 67.4% 
Asian/Pacific Islander  64.7%
African American/Black  72.1%
White  51.4%
Other  56.4%
Overall, half of voters (56.5%) chose to rank three candidates, while another 10.4% ranked two candidates and a third of voters (33.1%) chose one candidate.[18]  Nearly half (46%) of voters said ranking candidates was “easy” or “very easy”, and only 16% said it was either difficult of very difficult.[19]

Precinct analyses show a similar result. FairVote’s analysis of San Francisco’s 2004 District 1 Supervisor race showed that, on average, supporters of the two Asian American candidates (Lillian Sing and Rose Tsai) made more use of their rankings (2.56 and 2.58 respectively out of a possible three) than the supporters of the eventual winner, Jake McGoldrick (2.41).[20]  

This high use of rankings also was found by the Asian Law Caucus’s exit poll survey of the 2006 open seat race in District 4 (one of the city’s most heavily Asian districts). The ALC study found that 82% of Asian Americans ranked two or three candidates on their ballot, compared to 84% of non-Asian Americans.[21]  There were clear signs that voters used these rankings effectively. For example, even though voters were limited to three rankings, the race had four strong Asian candidates. Leading Asian organization split their endorsements among these candidates, raising a fear that Asian American voters would split their preferences in a race against a non-Asian candidate who was popular enough to finish second. But instead, the great majority of Asian voters created an ad hoc Asian coalition by choosing to rank three of the Asian candidates in some order. Indeed, a separate FairVote analysis showed that any one of these three Asian candidates would have defeated the non-Asian candidate in an IRV election because they would have picked up the second and third rankings from the supporters of the other Asian candidates.

F. Education campaign recommendations

The successful transition to IRV in San Francisco, as well as its successful use in other cities like Burlington (VT) and London (UK), indicates that any community can expect a similarly smooth transition with appropriate preparation. The most important factors boosting effective use of the IRV system are ones that can be done inexpensively: sensible ballot design, clear ballot instructions and well-trained pollworkers. In 2007 Cary (NC) demonstrated how inexpensive voter education campaigns can be, with the county board of elections spending less than $10,000 for voter education for the first IRV election in a city with more than 110,000 people. We examine two approaches to voter education.

San Francisco: Well-Funded Voter Education Campaign Option

With a relatively generous budget for voter education for its first IRV elections for seven of the city’s 11 wards during a high turnout presidential election, San Francisco in 2004 pursued an education campaign to voters and various communities that can serve as a model for a community wanting to allocate significant resources to voter education. The San Francisco Department of Elections launched a multi-pronged campaign that included print and electronic media, mass mailings, information in the Voter Information Pamphlet, and targeted outreach to specific communities. In addition, much thought went into ballot design and how to construct the precinct itself and train poll workers. This way, even a voter who had never heard of IRV would have a successful experience on Election Day—or when voting by absentee ballot, as more than three in ten voters have done in recent IRV elections in San Francisco.

A public relations firm was hired to create a clear and effective message to reach the voters. Posters, brochures, and public service announcements (PSA's) were used to explain IRV in several languages, including Spanish and Mandarin. To ensure that the voting precincts ran smoothly, the Department of Elections carefully trained its poll workers and ensured that voting instruction materials were easy to follow.[22]  The initial education campaign cost $1.70 per voter.[23]  In subsequent elections, the cost of the education campaign has been less expensive, costing approximately $0.50 per voter. Given that Burlington, VT and Cary, NC (see below) spent about $0.25 per voter for its first IRV elections, San Francisco spending should be viewed as an upper limit of a range of possible spending for a successful campaign.

A review of educational campaigns in San Francisco and other cities demonstrates several steps that are useful for successful and cost-effective voter outreach:

  1. Use a clear ballot design
  2. Draft clear voter instructions
  3. Set up a helpful, multilingual website with sample ballots to educate voters
  4. Allocate extra IRV training for poll workers
  5. Design the polling station with a goal toward educating voters
  6. Use media sources, including the ethnic media, effectively through PSA’s, press releases and advertisements
  7. The Voter Information Pamphlet should include a special section on IRV
  8. One or more separate citywide mailings to all registered voters
  9. Distribute helpful, language-appropriate brochures in public places like libraries and grocery stores.
  10. Encourage candidates to educate their supporters about IRV; it will benefit them as well.
  11. Concentrate education and outreach in the final month before voting takes place, as that is when the most voters are paying attention.
  12. Provide information in advance to reporters and editors so their stories have an educational component.

Cary in 2007: Modest Budget Voter Education Campaign.

In Cary’s city election for mayor, an at-large town council race and two district council seats, the Wake County Board of Elections in North Carolina spent less than $10,000 in a city of more 110,000 residents. Nineteen out of twenty voters reported that they understood the system, most of them well, and more than two-thirds preferred it to their old runoff system despite its adoption barely four months earlier. Voter education activities consisted of:

  1. The Board of Elections and town sent sample ballots in utility bills and issued a media advisory about the pilot and ballot change.
  2. The local newspaper ran articles about the new ballot design and pilot program.
  3. Board staff and volunteers visited civic organizations like the Kiwanis and American Legion to inform them of the new IRV ballot.
  4. Local radio stations ran 30-second PSA’s.
  5. The North Carolina Center for Voter Education created a video PSA for IRV, which ran on the government cable access station.
  6. The local League of Women Voters and other civic groups like Democracy North Carolina placed sample ballots at grocery stores, libraries and the Division of Motor Vehicles.

As with San Francisco, the most important step was a good ballot design that was intuitive to voters. There are increasing examples of best practices for such designs and instruction to assist other communities implementing IRV.

Instant runoff voting (IRV) has been used in a variety of locations around the United States and around the world. All available data resulting from exit poll surveys and precinct analysis show that the introduction of IRV has not hurt racial and ethnic minority voters, candidates or communities. These voters have understood IRV, preferred it to traditional two round runoffs and have successfully ranked their ballots, using them effectively. In San Francisco’s first citywide election with IRV, the six neighborhoods with the highest percentage of racial minorities enjoyed voter turnout four times higher than under the old two round runoff system.

In fact, much of the data shows that in some races IRV has been directly beneficial for racial minority voters and candidates by allowing them to use the ranked ballots to form ad-hoc voter coalitions. The effective use of rankings has allowed racial minority voters to reduce “split votes” among competing minority candidates, assuring that when these voters vote for their favorite candidate, that vote won't unintentionally lead to the election of their least preferred choice.

While there have been some variation in results among various racial and ethnic groups, those disparities have been minor and do not constitute disenfranchisement for any group. Rather, they pinpoint the areas where continued education will be helpful to assure that all voters continue to have a positive experience with instant runoff voting.

[1]Louisiana, South Carolina and Arkansas used traditional two-round runoff elections, but there is no time to mail the second round ballot to overseas voters. Hence, these states allow voters to rank a first, second and third choice and the runoff rankings are used to determine each voter's ballot in the second round, if one is necessary.

[2]SFSU IRV Survey, July 2006:, p. 25
[3]SFSU IRV Survey, July 2006:, p. 30
[4]Summary of several cities’ IRV exit polls,
[5]It is important to note that a voter who does not know anything about IRV can successfully rank candidates and cast a ballot by simply following the voting instructions. So even the approximate 13% or so of respondents who reported not understanding IRV should not be interpreted as meaning they did not effectively use their ballots or successfully rank their candidates. Many American voters do not understand how the Electoral College system works, yet still can cast an effective vote for president.
[6]SFSU IRV Survey, July 2006:, p. 10
[7]SFSU IRV Survey, July 2006:, p. 33
[11]Minimum share of votes, among all educational levels, that did not find the ballot confusing:
[12]SFSU IRV Survey, July 2006:
[13]SFSU IRV Survey, May 2005: SFSU IRV Survey 2004 Election.pdf, p. 4
[14]Christopher Jerdonek, Ranked Choice Voting and Voter Turnout in San Francisco Elections, p. 2
[15]Christopher Jerdonek, Ranked Choice Voting and Voter Turnout in San Francisco Elections, p. 5
[16]Christopher Jerdonek, Ranked Choice Voting and Voter Turnout in San Francisco Elections, p. 5
[17]Christopher Jerdonek, Ranked Choice Voting and Voter Turnout in San Francisco Elections, p. 9
[18]SFSU IRV Survey, July 2006:, p. 17. Note that because this race only had three candidates, a voter ranking two candidates in fact cast a fully effective ballot, because that voter’s ballot would never count for their third choice.
[19]SFSU IRV Survey, July 2006:, p. 21. It is important to note that this self-definition of “difficult” may not have an impact on voters casting an effective ballot. For example, some voters may have said that they found ranking difficult, not because the act of ranking was difficult, but instead simply because they considered all of the candidates other than their favorite to be unacceptable, or all candidates relatively similar. For example, a person making a list of movies to rent might say it was "difficult" to rank several movies in order of preference, but still be thankful for the opportunity to do so when finding that their favorite is not available.
[20]SFSU IRV Survey, May 2005: SFSU IRV Survey 2004 Election.pdf, p. 5
[21]Nov. 2006 Asian Law Caucus exit poll:
[22]New America Foundation: Description and Analysis of San Francisco's Voter Education and Outreach. (A fuller evaluation of the San Francisco Education and Outreach plan from the 2004 election may be found on the Web at
[23]City Clerk’s 2007 Municipal Elections After Action Report.