New York City held its first citywide ranked choice voting (RCV) election on June 22, 2021 in party primaries for mayor and all other city offices in the wake of 73% of voters backing a 2019 ballot measure proposed by a city charter commission. Key findings from FairVote's analysis include:
Eric Adams is projected to be the city’s second African American mayor, winning 51% of the vote in a close race against Kathryn Garcia and Maya Wiley. In the city council elections, candidates of color won 35 Democratic primaries, up from 26 primaries in the previous cycle. 29 out of 51 seats are projected to go to women in the general election, which will be the largest number of women in the history of the city council. Additionally, 6 seats went to LGBTQ+ candidates.
The candidate fields themselves were also historically large and diverse as candidates from a wide variety of backgrounds were empowered to run for office. 73% of all candidates were people of color and 43% were women or non-binary. Each race averaged a field of 6.5 candidates, over 2.5 times the 2017 average of 2.6. The introduction of RCV coincided with an expansion of public financing in New York City, possibly contributing to the large number of candidates running for office.
Several races resulted in trailblazing outcomes. As reported by Common Cause NYC, Crystal Hudson and Kristin Richardson Jordan will be the first out queer Black women on the City Council; Chi Osse, at 23, will be the youngest-ever Council member; Shahana Hanif will be the first Muslim woman and among the first South Asian Council members; Jennifer Gutiérrez will be the first Colombian-American; and Shekar Krishnan will be the first Indian American.
The Condorcet criterion states that the candidate who would win a one-on-one match-up against every other candidate should win the election. If such a candidate exists, they are known as the “Condorcet winner”.
In our cast-vote-record (CVR) analysis, we calculated head-to-head match-ups between candidates and found that every ranked choice winner was also a Condorcet winner, including the three candidates that "came from behind" to win their respective races.
In the Mayor's race, Eric Adams wins in head-to-head match ups against every other candidate, making Adams the Condorcet winner. He wins 50.45% of match-ups against Kathryn Garcia, who came in second place, followed by 54.89% of match-ups against Maya Wiley. This matches the order of elimination in the ranked choice tally, with Maya Wiley being eliminated in the penultimate round before Kathryn Garcia. However, interestingly enough, Kathryn Garcia falls short in a head-to-head match-up against Maya Wiley. This is because removing Eric Adams from the race reveals that his voters preferred Wiley to Garcia as their second choice, meaning Wiley wins 50.58% of match-ups against Garcia.
In 3 of NYC's 52 primary elections held with RCV, the ultimate winner was not the candidate who had the most first-choice support. These three "come-from-behind winners" demonstrated broad consensus in later rounds.
The come-from-behind winners are:
Come-from-behind wins typically happen about 10% of the time in competitive races, and they indicate a candidate who has built strong support among second- and third-choice preferences in addition to their first-choice support. Sometimes, come-from-behind winners indicate a situation which would have the potential for vote-splitting in plurality elections, but in which RCV identifies the candidate who is honestly preferred by more voters.
Inactive ballots occur when a ballot cannot be counted for a candidate in the current round of vote tabulation. Inactive ballots are sometimes called “exhausted ballots”. A ballot in NYC can become inactive in the following ways:
As is the case in other RCV jurisdictions, the vast majority of inactive ballots in New York City occurred due to voluntary abstention, or a voter choosing not to use all rankings. Pre-election polling suggested that most voters who planned not to fully rank their ballots say the primary reason is that they only liked one candidate.
The plots below show the average number of rankings used in each precinct, plotted against the portion of voters of each of five racial or ethnic groups in those precincts.
Upward-sloping lines indicate that when that racial group makes up a larger share of the population in a precinct, the average number of rankings used on ballots in that precinct also increases, suggesting members of that racial group ranked more candidates.
The two citywide contests for mayor and comptroller show different ranking behavior by demographic:
In the mayoral election, voters in heavily-White precincts tended to return more ballots with a high number of rankings, but the exact opposite pattern occurs in the comptroller election, the only other citywide election, where the most strongly positive relationship to highly-ranked ballots is precincts with large Black populations.
Given these divergent results, we conclude that ranking behavior is more dependent on the candidates and their campaigns rather than on a voter’s race or ethnicity. For example, previous FairVote work found that candidates who are perceived as front-runners and candidates who engage in anti-RCV rhetoric are likely to attract more “bullet votes”, or votes which select only that candidate and rank no others.
In the mayoral election, Eric Adams fits both of those criteria. Pre-election polling showed adams with a roughly ten-point lead and he made numerous public statements which were skeptical of RCV. Unsurprisingly, the many voters who chose Eric Adams as their first choice used the fewest average rankings among voters for the top five candidates.
Average Ranks from First-Choice Voters
In the comptroller's race on the other hand, there was less perception of a clear front-runner. Leading up to the election, Lander was trailing his opponent Corey Johnson in the polls. Additionally, Lander and Johnson both made positive statements about RCV prior to the election.
Given Adams’ strong support from Black voters in the mayoral race, the low ranking usage for Adams in particular may have contributed to the lower-than-average ranking usage for Black voters overall in that race, even while Black voters demonstrated the opposite ranking behavior in the comptroller’s election.
Below are the results for each of the Borough President races, five Democratic races and one Republican.
The borough president's race in the Bronx included 5 candidates and Hispanic or Latino voters have the strongest positive relationship with the number of rankings used, while White voters were less likely to use a high number of rankings. Three of the five candidates on the ballot were Hispanic or Latino.
AAPI voters in the Bronx also have a positive relationship, but there is more uncertainty given their smaller population in that borough, and the relationship falls short of statistical significance as determined by the p-value.
Brooklyn had the most competitive borough president race of the five boroughs, with 12 candidates competing. In Brooklyn, Black voters were particularly likely to use a high number of rankings on their ballots. A majority of candidates in Brooklyn were black.
In Manhattan, seven candidates competed to become borough president. Precincts with more Black voters cast more highly-ranked ballots, while precincts with larger AAPI populations tended to rank fewer candidates.
Five candidates competed in the Democratic primary in Staten Island. Black voters, Hispanic or Latino voters, and voters of mixed race or another race tended to rank more candidates.
Four candidates ran for the Republican nomination for Staten Island borough president, and precincts with more Black or Hispanic or Latino voters tended to cast more ballots with high use of rankings.
Queens had three candidates running for borough president, and precincts of all different racial / ethnic makeups appear to have engaged in similar ranking behavior.
Our analysis finds that ranking usage is dependent on the context of the election much more than by a voter’s race or ethnicity. We find that RCV allows every racial or ethnic group in this study to fully utilize their ballot, with no particular group disadvantaged by the change to a ranked ballot in 2021.
The mayoral race and several city council races were notable for candidates’ use of coalition-building RCV campaign strategies. In ranked choice voting, candidates are competing to be voters’ second and third choice as well as their first. This means that cross-endorsements can help candidates build coalitions outside of their traditional base to expand their outreach and improve their chances of winning. This is exactly what we saw when Andrew Yang endorsed Kathryn Garcia as a second choice in the mayoral race.
The Yang and Garcia coalition strategy came late in the election, after the start of early voting, but likely helped Garcia to pull ahead of Maya Wiley to earn a place in the final round where Garcia ultimately finished within one point of Eric Adams.
The impact of the coalition strategy is apparent when examining how ballots were transferred when Yang was eliminated in the sixth round.
Ballots Transferred to Finalist Candidates Upon Andrew Yang’s Elimination (Excluding Inactive Ballots)
https://e.infogram.com/68a925cf-eb33-4003-9d5a-9c2f10043e18?src=embedNYC yang-garcia alliance550334no0border:none;allowfullscreen
*Data averaged from two RCV polls: An Emerson Poll (in the field June 15-16), and a Citizen Data/Fairvote RCV Poll (in the field June 14-17).
Garcia and Yang made their first joint campaign appearance on June 19th. When comparing their polling averages prior to June 19 with the election results from June 22, Garcia increases her standing with Yang voters by nearly eleven percentage points. Meanwhile, Adams and Wiley each underperformed the polls’ predictions with Yang supporters.
While Garcia ultimately fell short, she and Yang demonstrated how cooperation, coalition, and asking for voters’ second and third choice votes can impact the outcome of an election.
Candidates in other New York City races also campaigned together or cross-endorsed each other. Several candidates in the Democratic primary for Richmond Borough President informally worked together, with candidate Cesar Vargas explaining, “I think ranked-choice voting has made (the campaign) more collegial… I can see my competitor not just as a candidate to compete against me, but also as someone who can also support me by working with his community because I can do the same for him, because at the end of the day he or she can be part of my voters’ options.”
Elizabeth Adams in council district 33 and Sandra Nurse in council district 37 both collaborated with other campaigns for training events or literature distribution and went on to win their elections.
Candidates who campaigned together or cross-endorsed but were not successful in their races include Rebecca Lamorte and Billy Freeland in district 5, Sara Lind and Jeffrey Omura in district 6, five candidates in district 7 who used the Twitter hashtag “#RankUs1Thru5”, Nabaraj KC and Austin Shafran in district 19, seven candidates in district 20 who united in an unsuccessful attempt to oust an establishment-backed candidate, Debra Markell and Harpreet Toor in district 23, and Scott Murphy and Andy Marte in district 34.
Cast Vote Records (CVRs) are digital records of how each voter voted. It is customary for RCV jurisdictions to release CVRs after RCV elections, which allow for independent verification and analysis.
The NYC Board of Elections posted CVRs here:
These files are clean human-readable files in .csv format, separated by each race on the ballot. FairVote produces this version of CVRs through our own software called the "RCV Cruncher". The "RCV Cruncher" software also performs other analysis based on the Cast Vote Records. See below for additional files.
* Because of formatting of the Board of Elections files, undervote counts are not guaranteed to be accurate.
** To mitigate the large size of these files, ballots with an identical ranking order have been combined into a single row. See the field titled "weight" for the number of observations of that ballot ranking order.
These tables show the round-by-round results for each election in .csv format. Ballots have been parsed using NYC's rules for overvotes and skipped rankings. Note that these files do not use batch elimination, allowing us to examine vote transfer at the individual candidate level.
Find out the average number of rankings used by voters in each contest, and whether ranking usage varied based on the voters' first-choice preference.
A table showing the results of head-to-head matchups between each pair of candidates. A candidate ranked higher than another candidate on a given ballot is assumed to win that voter's head-to-head matchup. All ranked candidates on a ballot are assumed to be preferred to all non-ranked candidates.
Read each table across rows. For example, in the file for the mayoral election titled "NewYorkCity_06222021_DEMMayorCitywide_percent.csv", read the first row with data as "Aaron Foldenauer wins 5.5% of matchups against Andrew Yang, 27.9% of match-ups against Art Chang, 11.0% of matchups against Dianne Morales," and so on.
A table showing how first-choice voters for each candidate had their ballots counted in the final round, including how many ballots counted for each finalist candidate and how many ballots became inactive because they did not rank either finalist.
A table showing the portion of ballots for each first-round candidate which ranked other candidates in their top three choices. These tables help us identify coalitions of voters when we see a high portion of voters for one candidate also ranking another candidate.
For example, in the file for the mayoral election titled "NewYorkCity_06222021_DEMMayorCitywide_percent.csv", we can see how Eric Adams voters ranked other candidates. 30.1% of Adams voters ranked Maya Wiley in their top 3 choices and 21.7% of Adams voters included Kathryn Garcia in their top 3 choices.