N.3 - RCV — Redondo Beach City Council, July 30, 2024

BLUE FOLDER ITEM
Blue folder items are additional back up material to administrative reports and/or public comments received after the printing and distribution of the agenda packet for receive and file.

CITY COUNCIL MEETING
JULY 30, 2024
N.3 DISCUSSION AND POSSIBLE ACTION TO PROVIDE DIRECTION TO STAFF RELATING TO THE DRAFT ORDINANCE AMENDING THE CITY’S MUNICIPAL

CODE RELATED TO RANKED CHOICE VOTING AS AN INSTANT RUNOFF MODEL.
• Sample Ballot 2023-03-07 – CA5 Instant Runoff Election • Resolution No. CC 2006-0610-099 – Break a Tie
• Public Comment
• Councilmember Nehrenheim - Executive Summary -
McCarty RCV Paper
• Councilmember Nehrenheim - Final RCV Study
RB-527 *
RB527*
PROPOSED CHARTER AMENDMENT MEASURE CA5

THE PEOPLE OF THE CITY OF REDONDO BEACH, CALIFORNIA, HEREBY AMEND SECTION 18.4 OF ARTICLE XVIII OF THE CHARTER OF THE CITY OF REDONDO BEACH TO READ AS FOLLOWS:

NOTE: New provisions or language added to the existing charter section are shown in BOLDFACE type; words and figures deleted from the existing charter section are shown in strikeout type.)
Sec. 18.4. Majority vote: Runoff elections. Instant runoff election.

A. Except as provided in this Section with regard to runoff elections, aThe City Council must by ordinance provide for an instant runoff voting system in the elections of each city elective office listed in Article VI, Article VIII, and Article X.

A majority ( more than half), of the votes cast for all candidates, or a majority of the operative votes on continuing ballots in subsequent rounds, is required for the election of a candidate to each such office as determined by an instant runoff voting system process detailed in the ordinance the council must enact. for each City elective office is required for the election of the candidate to such office. In the event no candidate for an elective office receives a majority of the votes cast for all candidates for such office, the two (2) candidates receiving the highest numbers of votes cast for any such office shall thereby qualify as candidates for such office at a runoff election to be held not later than seventy ( 70) days after such election. In the event that any person qualifies pursuant to law to become a write- in candidate for such office at the runoff election, the candidate who receives a plurality of all the votes cast for such office in the runoff election shall be elected.

B. Notwithstanding the foregoing, members of the Board of Education shall continue to be elected by a plurality of votes cast.

18.4—as added by election 11- 4-80; as amended by election 11- 6-84 and 3-7-95.

RB-528 *
RB528*
CITY ATTORNEY’S IMPARTIAL ANALYSIS OF MEASURE CA5
Background. The California Constitution gives charter cities such as Redondo Beach plenary authority” to provide for “ the manner in which, the method by which, the times at which, and the terms for which the several municipal officers ... shall be elected or appointed.” Currently Section 18. 4 of the Redondo Beach City Charter provides that:
a majority ( more than half), of the votes cast for all candidates for each City elective office is required for the election of the candidate to such office. In the event no candidate for an elective office receives a majority of the votes cast for all candidates for such office, the two (2) candidates receiving the highest numbers of votes cast for any such office shall thereby qualify as candidates for such office at a runoff election to be held not later than seventy 70) days after such election.

The above does not apply to candidates for members of the Board of Education who are elected by a plurality of votes cast.

The Measure. Measure CA5 (“ Measure”) was placed on the March 7, 2023 General Municipal Election Ballot by the Redondo Beach City Council. Measure CA5 would amend Section 18.4 of the Redondo Beach Charter to eliminate separate runoff elections and require the City Council by ordinance to provide for an instant runoff voting system in the elections of each city elective office.

When there are more than two candidates for one seat, an instant runoff voting system allows voters to rank candidates for that seat in order of preference when marking their ballots. Only one candidate can represent each of the voter ’s first, second, or third, etc., choices. For all city elective offices with more than two candidates, all first choices are initially tallied. If any candidate receives a majority of the first choices, that candidate is elected just as they would be under the current Charter Section. However, if no candidate receives a majority of first choices, the instant run-off ” process is triggered. The candidate receiving the fewest first choices is eliminated, and the voters for that eliminated candidate now have their second choices counted. The ballots are again tallied and the process continues until one candidate wins a majority. An instant runoff voting system is so named because it allows a majority winner to be determined in just one election. It is a substitute for a separate runoff election.

Measure CA5 is a charter amendment to be implemented by a later ordinance that the City Council would adopt, which will lay out the details of how the instant runoff voting system will work in the City of Redondo Beach. The ordinance will only affect elections for city office and will have no effect on ballot measures for any Federal, State, County, District or Board of Education elections. s/ Michael W. Webb
City Attorney
RB-529 *
RB529*
ARGUMENT IN FAVOR OF MEASURE CA5
Redondo Beach currently votes for it ’s elected officials in March elections using a traditional voting system. Whichever candidate gets more than 50% of the vote wins.

If no candidate receives more than 50% of the vote, a runoff is held in May. Runoff elections are expensive and have historically been hostile. The 2013 runoff election cost taxpayers nearly an additional $ 300, 000! Runoff elections also tend to have lower voter turnout than the City’ s March general election.

There is a better alternative election system— instant runoff voting. Using Instant Runoff, Redondo Beach will experience more amicable campaigns at a lower cost.

An instant runoff voting system determines a majority winner in a single, higherturnout election. The most popular instant runoff system is Ranked- Choice Voting RCV). With RCV, you rank candidates based on the order you prefer: 1st choice, 2nd choice, 3rd choice, and so on. If your favorite doesn’ t win, your vote still counts for your next candidate choice.

Voters using RCV in Oakland, San Francisco, San Leandro, Albany, Berkeley and Palm Desert say RCV is simple and they want to continue using it. An Instant Runoff Voting system encourage candidates to focus on issues instead of uncivil attacks on their opponents in hopes of becoming your second choice.

Measure CA5 was unanimously approved by the City Council.

We urge you to vote Yes on Measure CA5, the Redondo Beach Charter Change Amendment to allow instant runoff voting.

For more information, please visit www. RedondoIRV. org s/ Laura Emdee
Councilmember, District 5
s/ Todd Loewenstein
Councilmember, District 2
s/ Christian Horvath
Councilmember, District 3
s/ Nils H. Nehrenheim
Mayor Pro Tem
RESOLUTION NO. CC-061 0- 99

A RESOLUTION OF THE CITY COUNCIL OF THE

CITY OF REDONDO BEACH, CALIFORNIA,

PROVIDING FOR A PROCEDURE TO RESOLVE

TIE VOTES FOR ELECTIVE OFFICES AT

GENERAL MUNICIPAL ELECTIONS

CONSOLIDATED WITH THE BOARD OF

EDUCATION OF THE REDONDO BEACH UNIFIED

SCHOOL DISTRICT AND SPECIAL RUN OFF
ELECTIONS ( IF NEEDED).

WHEREAS, 915651 (a) of the Elections Code of the State of California authorizes the City Council, by majority vote, to provide for a procedure to resolve tie votes involving those candidates who received an equal number of votes at any Municipal Election held within the City of Redondo Beach; and WHEREAS, Section 18.4 of the Redondo Beach City Charter states that a majority ( more than half) of the votes cast for all candidates for Municipal Elective Office is required for the election of a candidate to such office; and WHEREAS, Section 18.4 of the Redondo Beach City Charter states that members of the Board of Education shall be elected by a plurality of votes casts.

NOW, THEREFORE, THE CITY COUNCIL OF THE CITY OF REDONDO BEACH, CALIFORNIA, DOES HEREBY RESOLVE AS FOLLOWS:

SECTION 1. That in the event no candidate for a municipal office qualifies by receiving more than half of the votes cast for that position and an equal number of votes are cast for two or more candidates qualifying for second place, the tie to decide the second place candidate for a run off election as dictated in Section 18.4 of the Redondo Beach City Charter shall be resolved as outlined in Exhibit " A" attached hereto.

SECTION 2. That in the event two or more candidates for a Municipal Elective Office in a run off election receive an equal number of votes creating a tie, the tie shall be resolved as outlined in Exhibit " A" attached hereto.

SECTION 3. That in the event two or more candidates for a position on the Board of Education receive an equal number of votes creating a tie for the qualified positions, the tie shall be resolved as outlined in Exhibit " A" attached hereto.

RESOLUTION NO. CC-0610-99

TIE VOTE RESOLUTION
PAGE NO. 1
SECTION 4. That the provisions of Sections 1, 2 and 3 shall apply at the next ensuing municipal election to be held in the City on March 6, 2007, and at each municipal election thereafter.

SECTION 5. That the City Clerk shall certify to the passage and adoption of this Resolution and enter it into the book of original Resolutions.

PASSED, APPROVED AND ADOPTED~;:

Mike Gin, Mayor
ATTEST:

STATE OF CALIFORNIA )

COUNTY OF LOS ANGELES) SS

CITY OF REDONDO BEACH )
I, Sandy Forrest, City Clerk of the City of Redondo Beach, California, do hereby certify that the foregoing Resolution No. CC- 061 0- 99 was duly passed, approved and adopted by the City Council of the City of Redondo Beach, California, at a regular meeting of said City Council held on the 3rd day of October, 2006, by the following roll call vote:

AYES: Aspel, Cagle, Szerlip, Diels, Parsons
NOES: None
ABSENT: None
ABSTAIN: None
est, City Clerk
APPROVED AS TO FORM:

Lv. tJ1j
Michael W. Webb, City Attorney
RESOLUTION NO. CC- 061 0- 99

TIE VOTE RESOLUTION
PAGE NO. 2
RESOLUTION NO. CC- 061 0- 99
EXHIBIT " A"
By Lot: The City Council shall forthwith summon the candidates who have received the tie votes, whether upon the canvass of the returns by the Councilor upon a recount by a court, to appear before the Council at a time to be designated by the Council. The Council shall, at that time and place, determine the tie by lot ( e. g., tossing a coin, etc.), or in order to maintain a sense of dignity in resolving a tie vote, Council may place each candidate' s name in an unmarked, sealed envelope. The Mayor will then select a person to draw an envelope and read the name of the winning candidate.

RESOLUTION NO. CC- 061 0- 99

TIE VOTE RESOLUTION
PAGE NO. 3
Measuring Public Opinion
The future of the galaxy is at stake. It seems like the frontrunners are either Han Solo or Luke Skywalker, but you prefer Princes Leia.

Who would you vote for in a traditional election?
How would you rank these candidates?
PROBLEM : VOTE-SPLITTING
Majority supported factions can split the vote
and end up electing a candidate the voters didn't support.

WHAT DO WE WANT IN A VOTING METHOD?
Simple: Easy to vote, easy to tally,
easy to audit. Transparent!
Honest: Safe to vote your
conscience. Incentivizes honest
voting.

Expressive: Voters are able to
show their full opinion.

Accurate: Winners reflect the will of
the people as best as possible.

Equal: Ensures an equally weighted vote
for all. Eliminates vote-splitting. Doesn't
give anyone an unfair advantage.

RANKED CHOICE DEAL BREAKERS
● Wasted votes: RCV wastes votes in multiple serious ways that are rare or non-existent in other methods.
● Centralized Tabulation: With RCV most rankings are not counted, leading to multiple serious logistical issues: centralized tabulation, delays, mistallies, and legal compliance and constitutionality issues. ● Inequality: RCV doesn't treat all voters equally, doesn't adequately address vote-splitting, and doesn't ensure voters an equally powerful vote.

Ranked Choice - Voided Ballot Issues
In STAR Voting voters can score candidates as desired
and their vote will be counted.

In RCV, the complex rules for voters lead to a high rate of "spoiled" or voided ballots.

No equal rankings allowed No skipped rankings allowed
Source: David Kimball. University of Missouri, St. Louis. Conference on Electoral System Reform. Stanford University. March 14-15, 2014. Voter Participation with RCV in the USA

RCV VOIDED BALLOT RCV VOIDED BALLOT
Studies have unanimously found that low income voters, voters of color, older voters, and non-English speakers have their ballots voided at much higher rates.

Implementation Logistics
● Not all rankings are counted
● Early returns can't be fully tallied
● Precinct level results aren't available
● Audits are prohibitively complex
● All ballot data is counted
● Early returns can be fully tallied
● Precinct level results are available
● Audits are simple
RCV ballots require
centralized tabulation
With STAR Voting
ballots are tallied locally

DETAIL OF EXHAUSTED BALLOTS
These ballots were not
counted in the deciding
round, despite being
numerous enough to have
flipped the election.

Analysis of full
candidate rankings
showed that Montroll
was actually preferred
over all others.

Wright lost, but his
voters never had their
2nd choices counted.

Kiss won, despite not being
the preferred candidate.

Inequality in RCV : Burlington
Kiss Wins Montroll was also the majority preferred candidate.

If all ballot data had been counted he would have won.

Tabulation Failures (Jurisdictions which mistallied and reported incorrect election results):
● NYC, New York mayoral election. Democratic primary 2021: It was discovered that 135k test ballots had been added to the official reported results. Board of Elections did not catch the error. ● Alameda County, California. 2022 General Election: In all races, the steps in RCV were conducted out of order, causing the wrong candidate to be certified as the winner in one race. Board of Elections did not catch the error.

Results Failures (Jurisdictions where RCV failed to elect the candidate preferred over all others):
● Alaska House Special Election, Aug. 2022. The candidate preferred over all others lost. The two Republicans split the vote and the seat flipped Democratic for the first time. ● Burlington, Vermont, 2009. The candidate preferred over all others lost. ● Moab, Utah, 2021. The candidate preferred over all others lost the first seat, though the election was multi-winner.

Implementation and Legal Failures (7 jurisdictions passed RCV but have been unable to implement it):
● Vancouver, WA. Santa Clara County, CA. Memphis, TN. Sarasota, FL. Ferndale, MI. Austin, TX. Hoboken, NJ Bans (10 states have now banned RCV):
● Tennessee, Florida, Idaho, Montana, South Dakota, Kentucky, Oklahoma, Alabama, Mississippi, and Louisiana. (A number of states have narrowly missed RCV bans that would have also banned alternative voting more broadly.) Repeals (Jurisdictions that implemented and then repealed RCV):
● Cary, NC. Aspen, CO. Ann Arbor, MI. Pierce County, WA. Sunnyvale, CA. Burlington, VT (later re-adopted). North Carolina. Hendersonville, NC. Eastpointe, MI. 10 cities in Utah.

REAL WORLD RCV FAILURES
5 = Best. 0 = Worst.

Vote your conscience.

Instructions:
• Give your favorite candidate(s) five stars.
• Give your last choice(s) zero or leave blank.
• Equal scores are allowed.
• Score other candidates as desired.

Results
The finalist with the
most votes wins!
★
Carmen
58%
Blake
37%
★ Equal Preference Vote: both finalists scored equally

COST AND IMPLEMENTATION
a) STAR Voting is compatible with existing election infrastructure and election integrity laws. Adding STAR to your vote tabulators just requires a software update and recertification. Not new hardware. b) Re-certification of hardware is not a barrier to implementation.

It's just a process. Certification is estimated at around 50k. c) Equal Vote Coalition has pledged to help ensure a comprehensive, inclusive voter education campaign, including offering staff hours, resources, and technical support.
d) The Equal Vote Coalition 501c3 has a "STAR Voting Implementation Fund" to help pay for certification and early implementation costs.

Which Animal Makes The Best Pet?
Scan the QR code to vote! equal.vote/pet
Try
Q & A

Ballot info is counted as the voter intended

STAR RESULTS
STAR Voting is tallied in 2 steps:

1.) Add up the stars. 2.) Add up the votes.

THE ORIGINS OF STAR VOTING
Score
Voting
Instant Runoff
(aka Ranked Choice)
In 2014, Equal Vote hosted a conference on electoral reform at the University of Oregon.

Advocates for the leading voting methods came together to present and discuss the options, but couldn't reach consesnsus. None of the proposals on the table could deliver on everyone's goals, and each proposal had its drawbacks.

This conversation sparked an idea. Could a hybrid proposal combine the best elements of previous proposals to deliver better results while addressing valid concerns with the older methods?

EMPOWERING VOTER VOICE
● Honesty is the best policy: Give 5 stars to your
favorite(s), and show your honest preference order
between the other candidates.
● Your vote automatically goes to the finalist you
prefer.
● If your favorite can’t win, your vote still makes a
difference.
● No matter what scores you gave to the finalists, your one full vote goes to the finalist you prefer.
● Allows voters to easily show both preference
order, and level of support.
● Best for cognitive load. Getting great data while
reducing voter errors and confusion.
● Votes are tallied with simple addition.

The 5 Star Ballot The Top-2 Runoff

FAIR - ACCURATE - EQUAL
● STAR tops the charts in every study and statistical analysis of voting method accuracy. ● The star ballot collects the best quality data possible on voter opinion, and then it uses all that data.
● No ballot data is wasted or ignored. Every ballot is counted in both rounds. ● STAR voting eliminates vote-splitting and the spoiler effect and guarantees that the voting method passes the highest bar for One-Person-One-Vote.

VOTING THEORY FACTS:
● The invention of STAR Voting was predicted in 2000 when studies on "Bayesian Regret" showed that Score voting when combined with a top two runoff was the best at electing the candidate who best represents the will of the people. ● The legal definition of one-person-on-vote requires ensuring an equally weighted vote when possible, which can only be done by eliminating vote-splitting.

Compare this with Instant Runoff Voting, which can actually incentivize Favorite Betrayal because, in order to pass Later No Harm, it ignores down-ballot voter preferences which could have been relevant. Despite the widespread claim that IRV eliminates the Spoiler Effect, Emily Dempsey (2018) demonstrates that in order to pass Later No Harm, a voting method must by definition fail to eliminate the Spoiler Effect and vote-splitting. For these reasons, we believe that the adherence to Later No Harm as a desirable pass/fail criterion is problematic.

In IRV, some voters whose favorites are eliminated will have their next choice counted, but voters whose favorites are eliminated in the final round will not. This biases elections against voters who prefer strong underdog candidates with broad support.

Counting the full ballot for some voters while ignoring relevant ballot data for others (as Later No Harm requires) gives voters a false sense of agency, may erode trust in the system and in voting reform in general, and is out of keeping with the spirit of one person, one vote.

Elections spoiled due to vote-splitting not only fail to elect the right winner, they also bias outcomes in predictable ways. “The Spoiler Effect occurs when a third candidate entering a race splits votes with a similar candidate who would otherwise win, thus causing a candidate less-preferred by the electorate to win instead.” Dempsey (2018) The center-squeeze Spoiler Effect in particular is pervasive in Plurality but is exhibited by IRV as well. When it happens, it fuels polarization and entrenches two party domination by preventing candidates in the middle of the field from winning." "In 1964, Wesberry v. Sanders, (Black, 1964) The U.S. Supreme Court declared that equality of voting—one person, one vote—means that “the weight and worth of the citizens’ votes as nearly as is practicable must be the same.” Passing the Equality Criterion ensures that it’s possible for voters who disagree to cast equally weighted and opposite votes, no matter how many candidates are on their side. Approval, Score, Smith/Minimax, and STAR Voting all pass this basic and ’practicable’ criteria; Plurality and Instant Runoff Voting do not." "The Later No Harm (LNH) criterion is effectively the opposite of Favorite Betrayal.

Saying nothing about whether it’s safe to vote for one’s favorite, LNH specifies that supporting other candidates in addition to a voter’s favorite cannot hurt their first choice.

This allows candidates to encourage supporters to rank others at no risk to themselves.

Both criteria are clearly desirable, but no deterministic voting method proposed to date has been able to satisfy both for elections with more than two candidates. We posit that rather than passing one but then failing the other criterion badly, voting methods should instead seek to maximize both. We believe that violations of LNH and FB, and their impacts on strategic voting incentives, should be evaluated statistically rather than with an axiomatic approach alone.

As we will show, STAR Voting incentivizes both honest and expressive voting by counting all ballot data given. The scoring round incentivizes voters to give their favorite(s) five stars. The runoff incentives voters to also give intermediate scores because showing honest preference order ensures their full vote will go to the finalist they prefer in the automatic runoff.

Peer Reviewed: STAR Voting, Equality of Voice, and Voter Satisfaction. Constitutional Political Economy. March 2023. https://starvoting.org/peer_review

COMBATTING STRATEGIC VOTING
● In STAR Voting the runoff creates strong incentives
for voters to show their preference order between the
candidates.
● Showing your preference order between the
front-runnners ensures that your voice makes a
difference in the runoff.
● In our current system, and Ranked Choice, it's
not necessarily safe to vote for your favorite.
● Voters don't want to waste their vote on a
candidate who can't win, so voting for the
"lesser of two evils" is common.
● In STAR Voting you should always give your
honest favorite 5 stars.

Ensuring that it's safe to
support your favorite
Incentivising voters to show
their preference order
VOTING THEORY FACTS:
● No voting method can eliminate strategic incentives in every scenario. ● No voting method can pass every desireable criteria.
● Many criteria are mutually exclusive, including "Favorite Betrayal" and "Later No Harm." ● The goal is to ensure good incentives and good outcomes in practice.

STAR Voting is constitutional all over the country, and beyond.

STAR Voting naturally complies with existing election laws, including one-person, one-vote, majority requirements, "most votes wins" clauses, local tabulation clauses, auditing requirements, and many more.

STAR IS NATIONALLY VIABLE
STAR Voting Endorsements
Comparing
STAR Voting
and RCV
(Ranked Choice Voting)
Instructions
Tabulation

COMPARING STAR AND RANKED CHOICE
WASTED VOTES - Ballot Limitations
Choice
● RCV ballots only allow
voters to rank a limited
number of candidates.
● Limiting the number
of ranks in RCV helps
prevent spoiled ballots,
but increases the
number of exhausted
ballots in races with
large fields of
candidates.
● With STAR, voters can
score as many or as
few candidates as they
want because equal
scores are allowed.

BULLET VOTING
Who are Bullet Voters?
● Voters who have a polarized opinion and only like one candidate. ● Voters who only have one candidate on their side.
● Lazy or rushed voters who don't take the time to vote expressively. ● Voters who strategically decide not to show support for other candidates, even though this is not a good strategy in either STAR or RCV.

In both Ranked Choice and STAR Voting some voters may "bullet vote" and only vote for their favorite. In both systems, if the voter did have a more nuanced opinion this is not effective and their vote is less likely to make a difference.

MEASURING PUBLIC OPINION
https://electowiki.org/wiki/2012_Occupy_Wall_Street_polls Note: Political leanings of participants are not expected to be representative of the general population. Votes (% of possible) RCV (first preferences)
Plurality
Legend:

Candidate
If there's a clear winner, RCV only counts voters 1st choices, so Plurality and RCV results looked almost identical and barely showed any support for the losing candidates. 40%
20%
0%
60%
80%

MEASURING PUBLIC OPINION
Candidate
Legend:

Votes (% of possible)
A 2012 exit poll in NYC looked at voter behavior under alternative voting methods; Plurality was compared to Approval, 5 Star, and Ranked Choice.

Obama won by a landslide with all methods, but 5 Star and Approval Voting results also showed support for the 3rd party candidates. 40%
20%
0%
60%
80%
Choose One
Approval
5 Star

More Info on
Ranked Choice
Voting (RCV)
WHAT IS RANKED CHOICE VOTING?
RCV Tabulation:
● 1st choice votes are counted and if a
candidate has a majority they are elected.
● otherwise, the candidate with the fewest
1st choices is eliminated and votes
transfer, if possible.
● In each round, your vote goes to the
remaining candidate you ranked highest.
● If your vote is unable to transfer, it is
discarded.

RCV TAKEAWAYS
● Voters rank candidates, but most
rankings are not counted. It depends
on the order of elimination.
● Selectively ignoring ballot data can
skew election results. Those whose
favorite is eliminated, but whose next
choice isn't counted are unfairly
disadvantaged.
● The candidate preferred over all
others can lose. Elections with three
or more competitive candidates are
the most likely to fail.
.

Implementation and Legal Failures (Jurisdictions that passed RCV but have been unable to implement it.):
● Vancouver, WA. Santa Clara County, CA. Memphis, TN. Sarasota, FL. Ferndale, MI. Austin, TX. Hoboken, NJ Bans (10 states have now banned RCV):
● Tennessee, Florida, Idaho, Montana, South Dakota, Kentucky, Oklahoma, Alabama, Mississippi, and Louisiana. (A number of states have narrowly missed RCV bans that would have also banned alternative voting more broadly.) Repeals (Jurisdictions that implemented and then repealed RCV):
● Cary, NC. Aspen, CO. Ann Arbor, MI. Pierce County, WA. Sunnyvale, CA. Burlington, VT (later re-adopted). North Carolina. Hendersonville, NC. Eastpointe, MI. 10 cities in Utah.

REAL WORLD RCV FAILURES
…
…
…
…
… …
…
RANKED CHOICE VOTING: Alaska US House '22 Special Election Takeaways:
● Palin was a 'Spoiler'. She split the Republican vote, causing them to lose. ● The Republican majority could have won if they had:
a. not run two candidates.
b. voted strategically for Nick Begich, the lesser evil. ● Ranking Palin 1st backfired and helped elect her supporters' last choice. If they hadn't voted at all, or had voted strategically, their 2nd choice would have won. ● Rather than electing the moderate from the majority faction, RCV fueled polarization by electing the minority faction candidate and flipping the seat.

Voters were wrongly told that:
a. it was safe to vote their conscience
b. their votes wouldn't be wasted
c. their 2nd choices would be counted if their first choice couldn't win d. the majority preferred candidate would win
e. RCV isn't polarizing
f. RCV eliminates the Spoiler Effect
These misleading claims and concerning real world case-studies have spurred a wave of RCV bans. Seven states have now banned Ranked Choice Voting statewide.

Advocates have to stop selling RCV with false claims!
At a glance:
● 60% voted for a Republican 1st choice.
● Nick Begich (R) would have defeated
Palin (R) or Peltola (D) head-to-head.
● Mary Peltola, the Democrat, won.
● 8% of votes were exhausted (not able
to be counted in the final round
between Peltola and Palin.)
Candidates:
● Nich Begich (R)
● Sarah Palin (R)
● Mary Peltola (D)
Vote totals*:
● 53% preferred Begich over Peltola
● 61% preferred Begich over Palin
● 51% preferred Peltola over Palin
* Not counting exhausted ballots.

Sources and additional references: electionscience.org/commentary-analysis/rcv-fools-palin-voters-into-electing-a-progressive-democrat/ arxiv.org/pdf/2209.04764.pdf Peer Review and Academic Articles on RCV
Ranked Choice was invented 150 years ago and
there is a wealth of data on where it delivers and
where it falls short.

RCV does well in races where only two
candidates are competitive, and successfully
eliminates "The Nader Effect" if a 3rd party
candidate is truly non-viable.

But, in elections with multiple viable candidates
Ranked Choice Voting breaks down, producing
non-representative and counterintuitive results.

For this reason RCV has not broken two party
domination in the countries where it's been
used the longest. RCV is not suitable for primary
elections or general elections with multiple
viable parties or candidates.
https://www.jstor.org/stable/2110786?seq=1
Abstract
It has long been recognized that Instant Runoff Voting
(IRV) suffers from a defect known as nonmonotonicity,
wherein increasing support for a candidate among a
subset of voters may adversely affect that candidate's
election outcome. The expected frequency of this type of behavior, however, remains an open and important
question, and limited access to detailed election data
makes it difficult to resolve empirically. In this paper, we develop a spatial model of voting behavior to approach
the question theoretically. We conclude that monotonicity failures in three-candidate IRV elections may be much
more prevalent than widely presumed (results suggest a
lower bound estimate of 15 % for competitive elections).

In light of these results, those seeking to implement a fairer multi-candidate election system should be wary of adopting IRV.
"Instant Runoff Voting (IRV) suffers
from a defect known as
nonmonotonicity, wherein increasing
support for a candidate among a subset
of voters may adversely affect that
candidate's election outcome"
https://www.jstor.org/stable/24507512
https://www.sciencedirect.com/science/article/abs/pii/S0261379414001395 Abstract
Some proponents of municipal election reform advocate
for the adoption of Instant Runoff Voting (IRV), a
method that allows voters to rank multiple candidates
according to their preferences. Although supporters
claim that IRV is superior to the traditional
primary-runoff election system, research on IRV is
limited. We analyze data taken from images of more
than 600,000 ballots cast by voters in four recent local elections. We document a problem known as ballot
“exhaustion,” which results in a substantial number of
votes being discarded in each election. As a result of
ballot exhaustion, the winner in all four of our cases
receives less than a majority of the total votes cast, a finding that raises serious concerns about IRV and
challenges a key argument made by the system's
proponents.
* Note: This study looked specifically at elections in which a majority was not found in the first RCV round of tabulation.
"Higher counts of overvotes were also
found, at times, among San Francisco
communities with more Latino residents
(Neely and Cook 2008), something shown
in a similar analysis of voters in Los
Angeles (Sinclair and Alvarez 2004), and in
areas with more foreignborn residents."
"What has not changed is the nature of the
discrepancies in who tends to overvote:
consistently, precincts where more
African-Americans reside are more likely to
collect overvoted, voided ballots. And this
often occurs where more Latino, elderly,
foreign-born, and less wealthy folks live.

The additional years of data show no
meaningful increase or decline in these
tendencies but rather bolster the earlier
study’s findings. In all of the elections we
examined, some voters were more at risk
than others of making disqualifying errors."
https://escholarship.org/content/qt8tm3s6hz/qt8tm3s6hz_noSplash_a5e40f23074e40a0b8a0be92279918ae.pdf "The controversy surrounding the 2000 U.S. presidential race fueled a variety of efforts to improve the administration of elections. Activists, benefiting from that momentum ... found some purchase at the local level in San Francisco, California. Proposition A passed in a 2002 March primary and replaced a two-round runoff system with instant-runoff voting (IRV).1 ... As the largest and longest-running application of IRV in the States, this serves as both a vanguard on the reform front and a test case for interested parties.2 "One concern in the discussion of any electoral reform is how well the public will understand a new system and what that implies for the equality of political voice. This is our focus. ... Concerns about the fairness of IRV led at least four jurisdictions to repeal similar reforms shortly after enacting them: Burlington, VT (2006–2009), Cary, NC (2007–2009), Pierce County, WA (2006–2009), Aspen, CO (2009).
"When we examined the 96 ranked-choice voting races in our sample from across the nation, our analysis found an average of

10.92 percent of ballots cast are exhausted by the final round of tabulation." "African Americans, Latinos, voters with less education, and those whose first language is
not English are more likely to be
disenfranchised with a ranked-choice voting
system."
When individuals leave columns blank on
their ballots and the candidate(s) they vote for
are eliminated from contention, their ballot is
not counted in the final tabulation… thereby
giving those who fully complete their ballot
more influence over the electoral process."
"only 50 percent of African Americans and 53
percent of Latinos ranked three candidates
whereas 62 percent of whites ranked a
candidate in all three columns."
https://mainepolicy.org/wp-content/uploads/RCV-Final-Booklet-.pdf "Too often, proponents of ballot initiatives advance
lofty claims to win support at the ballot box."
"In examining 96 ranked-choice voting races from
across the country where additional rounds of
tabulation were necessary to declare a winner, The
Maine Heritage Policy Center concludes that the
eventual winner failed to receive a true majority
61% of the time."
www.scribd.com/document/421886759/RCV-Final-Booklet#fullscreen&from_embed "the claim that ranked-choice voting always
provides a majority winner … is false and deserves
further scrutiny from voters."
"While candidates sometimes do receive a majority
of the total votes cast, a winner is often declared
only after a large number of exhausted ballots have
been removed from the final denominator."
Figure 5: Percentage of Competitive RCV Elections
That Did Not Result In A Majority Winner
60%
80%
40%
20%
0%

38.54%

61.46%
Winner Received More Than
50% of Total Votes Cast
Winner Received Less Than
50% of Total Votes Cast
Source: The Maine Herritage Policy Center
Minority Electorates and Ranked-Choice Voting
Executive Summary
Nolan McCarty1
January 10, 2024
Following the rise of political polarization in the United States and its negative consequences for effective governance, many scholars and activists have turned to electoral reform as a corrective.

Some arguments have focused on certain perceived deficiencies associated with the United States’ reliance on the combination of plurality voting and party primaries to elect most officeholders. The critics contend that such a combination leads to pathologies that reduce voter choice and generate polarized outcomes.

First, the critics argue that partisan primaries are dominated by the preferences of the most extreme voters who in turn nominate only extreme candidates leading to general elections in which voters must choose among polarized alternatives. Second, they argue that the use of plurality voting deters the entry of independent candidates and third-parties because they are very unlikely to obtain a plurality of the vote. Moreover, if such candidates do enter, they mostly serve as “spoilers” by elevating one candidate over another, often the less popular of the major party candidates. Finally, some scholars argue that the winner-take-all nature of plurality elections heightens conflict and negative campaigning in ways that heighten polarization.

To address these concerns, scholars and advocates have touted a set of reforms ranging from altering the rules for participation in primary elections, fusion voting, and various forms of proportional representation. But by far and away the most popular reform calls for the widespread adoption of ranked-choice voting (RCV). In a RCV election, voters rank a set of candidates, and election officials use these rankings to determine the election winners. In general, the system operates as follows:
• The ballot asks voters to rank the candidates in order of the voter’s preference. In some systems, they can rank all of the candidates while others ask only that the voters rank up to a certain number of candidates.
• The first stage of the vote tabulation procedure counts the first rank votes. If any candidate receives a majority of the first rank votes, she wins the election. This is often referred to as the “first round.”
• If there is no majority winner in the first round, the last place candidate, as well as any additional candidates that have been mathematically eliminated, are eliminated from the tabulation, and the vote tabulation proceeds to a “second round.” 1 Susan Dod Brown Professor of Politics and Public Affairs, School of Public and International Affairs, Princeton University, [email protected]. This research is partially supported by the Center for Election Confidence.
• In the second round, the votes are recounted using the first ranked votes of the remaining candidates and the second ranked votes of those who supported one of the eliminated candidates.

If a candidate obtains a majority on this round, she is the winner. If not, the process continues to a third round, and so on, with candidates being eliminated in each round while counting first, second and third or subsequent ranked votes, until there is a winner receiving a majority of first rank votes plus subsequent ranked votes. • Significantly, if a voter’s first rank is for a candidate who is eliminated, but does not rank a candidate still standing in a subsequent round, that voter’s ballot is “exhausted” and eliminated from subsequent tabulation rounds.

Advocates of RCV suggest many ways in which it would improve upon plurality voting. First, the promoters contend that the system improves the electoral fortunes of small parties and independent candidates. By encouraging a larger set of candidates to contest office, advocates argue that RCV provides voters with more choice, which in turn should result in greater voter turnout and engagement. RCV also purportedly eliminates the possibility of spoiler candidates who siphon off too many votes from the most popular major candidate. Finally, advocates argue RCV elections are more legitimate because the winner has earned the support of a majority of the electorate.

As discussed in my research paper, the evidence that RCV has demonstrated these advantages in practice is mixed, at best. But even if RCV elections achieve the touted features, RCV raises important questions about the impact on minority and disadvantaged voters and their opportunities for electoral representation and effective influence in election outcomes.

First, RCV elections, by design, are more majoritarian than plurality elections. Indeed, a common argument is that they provide a way to get the benefits of a majority-runoff system without holding separate elections. Whereas minority candidates have some chance of winning plurality elections when votes are closely divided between majority-group candidates, RCV works to reduce those opportunities especially in the presence of racially or ethnically polarized voting. RCV advocates may defend that attribute by arguing that RCV would increase the likelihood that the majority-candidate most favorable to the minority group wins. But such an outcome depends on majority candidates willing to appeal to minority voters to obtain their second-preference votes and for the minority voters to fully use their ballots to support such candidates in the later rounds. However, if the preferences of the majority and minority groups are sufficiently distinct, both majority candidates may compete for the support of majority voters in hopes of entering a second round against the minority-backed candidate in which they will win. Moreover, minority-group voters may “exhaust” their ballots and fail to rank either majority-backed candidate. Thus, such voters lose influence in the case of a majority-versusmajority second round. Such arguments highlight the problems of ballot exhaustion which previous research has found to be both endemic to RCV and concentrated in minority electoral precincts.

Given these concerns, my paper focuses on the concentration of exhausted ballots among minority electorates as well as its potential to decrease representation and electoral influence of minority voters. Specifically, I examine two cases in which RCV was recently adopted and show that exhausted ballots were far more common in precincts and electoral districts with high concentrations of minority voters. The first case is the New York City Democratic Primary elections held in 2021. Using micro-data on cast vote records (CVR) combined with demographic records of primary voters from the voter registration file, I show that electoral districts with large concentrations of minority voters cast substantially more exhausted ballots than other districts. These correlations persist even in those cases where a co-ethnic candidate advances to the final round of tabulation. I then examine the introduction of the Top Four Primary system and RCV general election in Alaska. There I combine the cast vote records with Census demographics to examine the correlates of ballot exhaustion in the special and regular elections held there in 2022. I find that areas with high concentrations of Native Alaskans are more prone to ballot exhaustion.

Key Findings
• Except in cases where a co-ethnic reached the final round of tabulation, I find that exhaustion rates in the NYC Democratic primaries for executive office were higher in precincts with high concentrations of minority (Black, Asian, and Hispanic) primary voters than they were in predominately White precincts. • Exhaustion rates in the NYC Democratic mayor’s primary were lower for Black electorates than White electorates. But this was the result of high Black support in firstrank preferences for Eric Adams who proceeded to the final round and won the election.

Exhaustion rates in Black precincts were higher than those of White precincts among those voters who did not rank Adams first.
• In the executive office primaries, the proportion of voters ranking only a single candidate was generally higher in minority precincts.
• In the Democratic primaries for City Council, ballot exhaustion was as high as the rates of wasted votes (votes for non-competitive candidates) in the plurality elections of 2017.

Moreover, the 2021 RCV elections had fewer majority winners and had lower winning margins than 2017 plurality elections. Both of these outcomes run contrary to the arguments proffered by RCV advocates. These patterns can be explained primarily by the vast proliferation of candidates under RCV.
• In the Council elections, exhaustion rates were considerably higher in minority precincts than in predominately White ones. These disparities were lower, however, when there was a co-ethnic candidate in the final round. The racial and ethnic disparities were greatest when there were a large number of candidates. This result suggests that minority electorates were less able to take advantage of the expanded “choice” of candidates. • Ballot roll-off (also known as “drop-off” in down-ballot races) between the mayor and Council primaries was much larger in 2021 under RCV than in 2017 under plurality.

This is contrary to claims that RCV would boost voter engagement. Moreover, roll-off rates tended to be the highest in minority precincts, especially when there was not a strong co-ethnic candidate.
• The results in Alaska largely confirm those of NYC for heavily Alaskan Native precincts.

Their exhaustion rates were higher in all state-wide races and for state legislative races except in the case of the U.S. House election which featured a co-ethnic winner.

Conclusions
Partisan polarization and related political dysfunctions have greatly increased interest in fundamental reforms to our electoral processes. But such reforms often come with a variety of trade-offs and unintended consequences. Therefore, careful scrutiny and evaluation of the effects of reforms is crucial.

Ranked choice voting is clearly a reform that has excited a lot of people. Its advocates suggest that it can both turn down the temperature surrounding electoral politics and increase the diversity of choices available to voters. But scholarly attempts to evaluate such claims and to root out the downsides are still in their infancy. In my paper, I try to evaluate one such downside -- the high rates of ballot exhaustion and their concentration in precincts with large minority populations. The findings suggest that these are indeed drawbacks of RCV. Across a variety of electoral contexts in New York City and Alaska, I find consistent correlations between the ethnic and racial composition of a precinct and the share of exhausted ballots. These correlations are especially large when there are large numbers of candidates and when there are not strong coethnic candidates in the race.

These findings are consistent with RCV providing an advantage to majority-group voters over minority-group voters. Whereas RCV allows majority-group voters an additional opportunity to resolve candidate coordination problems, the patterns of ballot exhaustion suggest that minoritygroup voters are not taking full advantage. Whether those higher rates of exhaustion are due to ballot complexity, lower levels of information and mobilization, or racial and ethnic polarization, it is clear that the potential effects of RCV on minority voters needs to be carefully scrutinized before adoption.

Minority Electorates and Ranked Choice
Voting
Nolan McCarty*
January 10, 2024
Abstract
Ranked Choice Voting (RCV) has become a very popular reform designed to mitigate several of the perceived flaws inherent in single-district plurality electoral systems.

However, relatively little attention has been paid to how RCV might impact the representation and influence of minority voters. In theory, RCV poses several difficulties for minority representation. First, RCV is by design a majoritarian system in that the winner must claim the support of a majority of the participating voters. Thus, RCV forecloses opportunities under plurality voting for minority-group-backed candidates to win elections when majority-group voters fail to coordinate on a single candidate.

Second, this problem is compounded to the extent that majority-group candidates lack incentives to appeal to minority-group voters. Such incentives will be lacking to the extent to which minority voters are unwilling or unable to rank multiple candidates. So patterns of ballot exhaustion and truncation across demographic groups is key to understanding how RCV might affect opportunities for minority-group voters.

Third, high ballot exhaustion rates among minority-group voters would mean that those voters exercise less electoral influence. In this paper, I examine the racial and ethnic patterns of ballot exhaustion in the 2021 New York City Democratic Primary and the 2022 elections in Alaska. I find strong evidence that electorates with heavy concentrations of ethnic and racial minorities have substantially higher rates of ballot exhaustion. These findings raise important questions about the impact of RCV on the electoral influence of such groups.
*School of Public and International Affairs, Princeton University, [email protected]. This research is partially supported by the Center for Election Confidence. I would also like to thank Eric Manning and the data Driven Data Science Initiative for assistance in navigating the voter files. I also thank Roel Bos, Peter Buisseret, Dahyun Choi, Alina Dunlap, Kyuwon Lee, John Kastellec, Conner Phillips, Anthony Taboni, and Hye Young You for helpful comments. 1 Introduction
Following the rise of political polarization in the United States with its negative consequences for effective governance, scholars and activists have touted electoral reform as a possible solution.1 These arguments have focused on perceived deficiencies associated with the United States’ reliance on the combination of plurality voting and party primaries to elect most officeholders. According to the critics that combination leads to pathologies that reduce voter choice and generate polarized outcomes. First, the critics claim that partisan primaries are dominated by the preferences of the most polarized voters who in turn nominate only extreme candidates leading to general elections in which voters must choose among polarized alternatives. Second, they argue that the use of plurality voting deters the entry of independent candidates and third-parties because they are very unlikely to obtain a plurality of the vote. Moreover, if such candidates do enter, they mostly serve as “spoilers” by elevating the less popular of the major party candidates. Finally, some scholars argue that the winner-take-all nature of plurality elections heightens conflict and negative campaigning in ways that exacerbate polarization.

To address these concerns, observers have suggested a set of reforms ranging from altering the rules for participation in primary elections, fusion voting, and various forms of proportional representation. But by far and away the most popular reform calls for the widespread adoption of ranked-choice voting (RCV). In a RCV election, voters rank candidates in order of preference, and election officials use these rankings to determine the election winners. Although advocates argue for a wide range of procedures, actual implementations of RCV tend to operate as follows:2

1. The ballot asks voters to rank the candidates in order of preference. In some systems, voters may rank all of the candidates while others systems ask only that the voters 1For discussions of polarization and electoral reform McCarty (2019) and Drutman (2020). 2See Santucci (2021) for a discussion of the procedural variants. 1 rank up to a certain number of candidates.

2. The first round of the tabulation procedure, counts the first-rank votes. If any candidate receives a majority of the first-rank votes, she wins the election. If no candidate secures such a majority, the last place candidate, as well as any other candidates that have been mathematically eliminated, are dropped, and the tabulation proceeds the second round.

3. In the second round, the votes are recounted using the first-ranked votes of the remaining candidates and the second-ranked votes of those who supported one of the eliminated candidates. Those voters who only ranked one of the eliminated candidates are said to have “exhausted” their ballots.

4. If a candidate obtains a majority of the votes from non-exhausted ballots in this round, she is the winner. If not, the process continues through multiple rounds where votes are tabulated based on each voter’s highest ranking of the non-eliminated candidates. A winner is declared once a candidate obtains a majority of the votes from the non-exhausted ballots.

Advocates of RCV tout many ways in which it would improve upon plurality voting.

First, the promoters contend that the system improves the electoral fortunes of small parties and independent candidates. By encouraging a larger set of candidates to contest office, advocates argue that RCV provides voters with more choice, which in turn should result in greater voter turnout and engagement. RCV also purportedly eliminates the possibility of spoiler candidates who siphon off so many votes that the most popular major candidate loses. Finally, advocates suggest RCV elections are more legitimate because the winner is required to earn the support of a majority of the electorate.

As discussed below, the evidence that RCV has demonstrated these advantages in practice is mixed. But even if RCV elections achieve the promised features, their usage raises 2 important questions about the impact on minority and disadvantaged voters and their opportunities for electoral representation. First, RCV elections are intended to be more majoritarian than plurality elections. Indeed, a common argument is that they provide a way to get the benefits of a majority-runoff system without holding separate elections.3 Whereas minority candidates have some chance of winning plurality elections when there is a close split in the vote among majority-group candidates, RCV works to reduce those opportunities especially in the presence of racially- or ethnically-polarized voting. RCV advocates defend that attribute by arguing that RCV would increase the likelihood that the majority-candidate most favorable to the minority group wins. But such an outcome depends on majority candidates’ willingness to appeal to minority voters to obtain their second-rank votes and for the minority voters to fully use their ballots to support such candidates in the later rounds. But recent theoretical work by Buisseret and Prato (2022) casts doubt that such conditions hold generally. In their model, if the preferences of the majority and minority are sufficiently distinct, both majority candidates may compete for the support of majority voters in hopes of entering a second round against the minoritybacked candidate in which they will win. Moreover, minority group voters may exhaust their ballots and fail to rank either majority-backed candidate. Thus, minority voters may actually lose influence in the case of a majority-versus-majority second round. Such arguments highlight the problems of ballot exhaustion which previous research has found to be both endemic to RCV and concentrated in minority electoral precincts.4 In this paper, I contribute to the literature on RCV and ballot exhaustion by focusing on its disproportionate concentration among minority electorates as well as its potential to decrease representation and electoral influence of minority voters. In section 2, I sketch an 3For this reason, some advocates refer to the system as Instant Runoff Voting. 4On the magnitude of ballot exhaustion, see Burnett and Kogan (2015) and McCarty (2020). On the correlations with race and ethnicity see Polubinksi (2023) and https://www.politico.com/states/new-york/city-hall/story/2021/09/08/ lower-income-areas-of-nyc-had-a-harder-time-with-ranked-choice-voting-1390719. 3 argument that RCV systems can be harmful to minority voters especially when those voters are more likely to exhaust ballots. I then examine two cases in which RCV was recently adopted and show that exhausted ballots were far more common in precincts and electoral districts with high concentrations of minority voters. The first case is the New York City Democratic Primary held in 2019 which is discussed in Section 3. Using micro-data on cast vote records (CVR) combined with demographic records of primary voters from the voter registration file, I show that electoral districts with large concentrations of minority voters cast substantially more exhausted ballots than other districts. These relationships are especially strong in heavily Hispanic and Asian districts. These correlations persist even in those cases where a co-ethnic candidate advances to the final round of tabulation.

In section 4, I examine the introduction of the Top Four Primary system and RCV general election in Alaska. There I combine the cast vote records with Census demographics to examine the correlates of ballot exhaustion in the special and regular elections held there in 2020. Given the very low numbers of Black and Hispanic voters (and the political salience of those identities), I do not find strong correlations between exhaustion and the size of those populations. I do find, however, that areas with high concentrations of Alaskan Natives are more prone to ballot exhaustion. I conclude and discuss the implications of the results in section 5.
2 Theoretical Framework
To illustrate how RCV may negatively impact minority voters, consider a hypothetical example where there are three candidates representing three blocs of voters.5I label the candidates/blocs as A, B, and C. Further, I assume that blocs A and B are part of the “majority” group and C is the “minority.” Consequently, the total number of voters 5See Buisseret and Prato (2022) for a more fully-elaborated model along the lines of this example. 4 in A and B is larger than the number of voters in C.
6 However, bloc C may be larger
than either bloc A or B individually. Moreover, I assume that no single bloc constitutes a majority of the electorate. Assuming that voters in each bloc vote sincerely for the candidate representing them, plurality elections elect the candidate of the largest bloc.

Voters may, however, choose to vote strategically. If majority-group voters are able to coordinate electoral support for candidate A or B, then that candidate wins. If they fail, however, the minority candidate C may win.

Now consider the effects of ranked choice voting. As a first case, assume that C is the largest single bloc and would therefore win a first-past-the-post election. Without loss of generality, let A be the second largest bloc. Under the assumption that all voters give the highest rank to the candidate of their bloc, C does not win in the first round. Moreover, the winner is determined by how voters in B allocate their second ranks between A and C.

If B prefers A as we might expect on shared in-group status, then A wins. Thus, RCV has insured that majority voters can coordinate on a single candidate, making it impossible for the minority candidate to win.7
As a second case, assume that C is the smallest group. Consequently, the second round is likely to be a contest between the majority candidates A and B. C’s only opportunity for influence is when its second-place ranks to help elect the majority candidate most sympathetic to its interests. Ideally, the majority candidates would appeal to C’s voters to enhance their electoral prospects. But suppose C’s voters “exhausted” their ballots by failing to rank A or B. This behavior would both reduce the likelihood that the most sympathetic majority candidate would win as well the incentives for either majority candidate to compete for C’s support. The C voters will, therefore, have less influence if their exhaustion rates are high.
6Depending on the context, majority and minority may be designated based on racial, ethnic, socioeconomic or partisan categories. 7See Morton and Rietz (2007) for a similar argument and consistent experimental evidence in the context of traditional runoff elections.
5 These two examples illustrate how minority electorates may be negatively impacted by the adoption of ranked-choice voting. First, RCV helps to ensure that majority voter blocs can coordinate on a single candidate thereby depriving minority voters of the ability to elect a candidate of choice even in cases where the minority bloc is the largest. Second, RCV may dilute minority voter influence to the extent to which those voters exhaust their ballots by failing to rank the majority-group candidates.

As the second example shows, the impact of RCV on minorities depends on the extent to which they disproportionately cast “truncated ballots” (i.e., ballots that doe not rank all alternative, viable candidates) that lead to their non-participation in the final round of tabulation. Consequently, much research on RCV focuses on truncation and ballot exhaustion, see e.g. Burnett and Kogan (2015). In a study of municipal elections in three cities, Polubinksi (2023) finds strong evidence that rates of ballot exhaustion are higher in precincts with heavy concentrations of minority voters. 8 Other scholars have noted that minority electorates have faced other challenges in using RCV. Neely and Cook (2008) and Neely and McDaniel (2015) find that rates of RCV overvoting, applying the same rank to multiple candidates, was significantly higher in minority precincts in San Francisco municipal elections. More recently, Cormack (2023) finds that ballot disqualifying overvoting was higher in non-White assembly districts in the 2021 NYC mayoral primary, and that correlation can be attributed to lower incomes and educational attainment.

Perhaps the most important question for my purposes is why voters would choose 8Survey experimental work has found less evidence for racial and ethnic disparities in ballot truncation or exhaustion. But these studies may suffer a lack of external validity. For example, Coll (2021) asks respondents to rank the top five 2020 Democratic presidential candidates and finds that age and gender are the only demographic categories correlated with truncation. But this setting may not generalize for a variety of reasons. First of all, the 2020 primary was a very high profile election where voters could be assumed to have reasonably good information about a wide variety of candidates – far different than the elections analyzed in this paper. Second, as discussed below, the presence of strong minority candidates like Kamala Harris and a White candidate with strong endorsements from Black elites, Joe Biden, can affect patterns of truncation and exhaustion relative to elections without such candidates. 6 not to fully participate by truncating their ballots. Voters who truncate their ballots are deprived of potential opportunities to influence the election outcome, even though their preferences among lower ranked candidates may be decisive. In fact, if the goal of a voter is to maximize electoral influence, she should always rank at least n − 1 candidates in an n candidate election.9
Underscoring this problem, Kilgour, Gr´egoire and Foley (2020) conduct a simulation study of the consequences of truncated ballots in RCV elections. The authors find that ballot truncation is very common and hard to rationalize. But more importantly, the authors demonstrate how some of the supposed salutary properties of RCV fail when voters do not fully participate. First, they find that even small amounts of truncation can alter the identity of the election winner, especially in elections with more than three candidates.

Often these distortions disadvantage and result in outcomes that are contrary to the will of the voter whose ballot is truncated.

Given the lack of a clear strategic motive, there are at least two other explanations for why voters might truncate ballots. The first is that voters are voting expressively and refuse to rank any candidates but their top choice. The most plausible form of such behavior is based on partisanship or group-identity – voters refuse to rank candidates from their out-party or group. While such behavior may be intentional and well-informed, it does undermine the stated objectives of RCV elections. Low rates of full participation make it less likely that the winner is supported by a majority of the total electorate, more likely that a candidate can be a spoiler, and less likely that a Condorcet winner is victorious. Moreover, persistence of this form of expressive partisanship and/or identity politics suggests that RCV is not succeeding in building cross-party/group coalitions and reducing polarization. Specifically, racial polarization may manifest as ballot truncation 9Unless of course the ballot design limits the number of candidates that can be ranked. Such limitations are commonly imposed for the convenience of electoral authorities, but they obviously exacerbate the problem of ballot exhaustion. Below I present evidence that such limitations have only a modest effect on ballot exhaustion.
7 and exhaustion. Most importantly, as discussed above, it may dampen the incentive for candidates to appeal to those groups who tend to exhaust their ballots. Moreover, as candidates reduce their appeal to high exhaustion groups, members of those groups have even less incentive to fill out complete ballots.

A second source of ballot truncation and exhaustion is that voters have poor information about how RCV tabulation works or lack adequate information about candidates necessary to do a full ranking. That truncated votes count less than fully participating ones suggests the possibility that low information voters may be relatively disadvantaged by RCV. Such variances in participation rates based on information gaps are troubling. And if those information disadvantages are concentrated on certain communities on the basis of race, ethnicity, education, or socio-economic status, the democratic legitimacy of RCV might be called into question.

My paper does not address other potential effects of RCV on minority electoral success and representation. But the prior literature suggests few benefits. McDaniel (2016) finds that turnout dropped in San Francisco mayoral elections following the adoption of RCV, especially among African-Americans and Asians. In a study of several RCV cities matched against comparable plurality cities, Kimball and Anthony (2016) find a four percentage point drop in turnout associated with RCV, although the estimate is not statistically significant on its own. In a more recent study, McDaniel (2019) finds a statistically significant five percentage point drop due to the introduction of RCV in municipal elections relative to similar cities that maintain plurality electoral systems. While there is disagreement about the magnitude and statistical reliability of the estimated declines in voter turnout, I am not aware of any study that finds a boost in turnout associated with switching to RCV from plurality voting.

I also do not consider whether RCV increases or decreases the salience of racial and ethnic differences. But McDaniel (2018) finds indicates that RCV did little to reduce – 8 and may have increased – racially polarized voting in mayoral elections. Finally, I do not consider effects of RCV on the pool of minority candidates, but note that Colner (2023) finds a modest, but short-lived, boost in the diversity of the candidate pool upon adoption of RCV. Experimental work reported by Santucci and Scott (2021) finds that RCV does not increase minorities’ interest in running for office. More generally, Vishwanath (2022) finds that RCV has little effect on the ideological composition of the candidate pool. 3 New York City Democratic Primary
After several years of study, in 2019 the New York City Charter commission proposed a voter referendum calling for the use of RCV voting in NYC municipal elections. The referendum was supported by 75% of those voting in a low turnout affair. The most direct motivation for the adoption of RCV was the elimination of costly runoff elections necessitated by the NYC Charter’s requirement for plurality winners with at least 40% of the votes cast. But advocates touted many other claims about RCV’s ability to produce majority winners, lessen negative campaigning, and promote moderation. To the extent that there was opposition, however, it focused on the possible negative consequences for minority and poorer voters.10
The first use of RCV voting in NYC was for the city-wide primary elections held on June 22, 2021. Those elections nominated candidates for Mayor, Comptroller, Public Advocate, the five Borough Presidencies, and the City Council. Although RCV was used in the Republican and Conservative party primaries, I focus on the Democratic primary. Table 1 reports some basic information about the outcomes of those contests. The table illustrates several important points.

The first is that even in elections with large numbers of candidates, the leading candidate 10See https://www.nytimes.com/2019/10/28/nyregion/ranked-choice-voting-ny.html and https://ibo.nyc.ny.us/cgi-park3/2019/10/22/eliminate-the-need-for-citywide-run-off-elections/. 9 Table 1: 2019 NYC Democratic Primary Outcomes
Round 1 Final Share %
Election # Cands Winner Runner-Up Plurality Share of Total Rounds Exhausted Mayor 13 Eric Adams Kathryn Garcia 30.7 50.4 42.9 8 14.9 Comptroller 10 Brad Lander Corey Johnson 30.9 51.9 39.3 10 24.4 Public Advocate 3 Jumaane Williams Anthony Herbert 70.0 70.0 70.0 1 0.0 Manhattan President 7 Mark Levine Brad Holman 28.3 53.8 42.4 7 21.2 Brooklyn President 12 Antonio Reynoso Jo Anne Simon 27.9 54.9 37.3 11 32.0 Queens President 3 Donovan Richards Elizabeth Crowley 41.1 50.3 47.2 3 6.2 Bronx President 5 Vanessa Gibson Fernando Cabrera 39.5 53.5 48.7 3 9.0 Staten Island President 5 Mark Murphy Lorraine Honor 46.5 65.0 57.5 6 11.5 received a very large vote share. Only in two cases did the leading candidate fail to get 30% of the vote. In three cases, candidates received more than the 40% threshold that would have avoided a runoff under the old system. In every case, the first round plurality winner went on to win after the RCV tabulations were conducted.11 A second notable finding is the large number of exhausted votes. With the exception of the Public Advocate’s race which was decided in a single round, the rates of exhaustion are very high. In a recent report, McCarty (2020) shows that the average rate of ballot exhaustion across 95 municipal RCV elections is 10.5%. Therefore, five of the seven contests with RCV tabulations were well above the historical norm. Moreover, exhaustion may have been consequential in some cases. In six of these elections, the percentage of exhausted votes exceeded the margin between the final two candidates. This fact implies that had there been fewer exhausted ballots, the results may have changed. This pattern also implies that in those elections, the winner failed to obtain support from a majority of voters who cast ballots in the election (see “Share of Total” column). In two cases, the Comptroller and Brooklyn President, the winner did not even crack 40% of the votes cast, the old threshold for triggering a runoff election under the Charter. These outcomes are troubling given that RCV is touted as producing winners supported by a majority of the electorate.12 11In the case of public advocate, Jumaane Williams won in the first round, therefore avoiding RCV tabulation.
12It might be tempting to blame these high exhaustion rates on the fact that the ballot only allowed voters to rank up to five candidates. But note that there are minority winners in two races with five or fewer candidates – the Bronx and Queens presidencies. Moreover, in the Mayor’s race, the average exhaustion rate of those who ranked five candidates was only 6.4% compared to 21.8% for those who ranked fewer.

So exhaustion rates would have been high even without the five candidate limitation. 10 Table 2: 2019 NYC City Council Primary Outcomes
Avg One Multi RCV Sub-Majority Avg Exhaust Avg Exhaust
Number Cands Round Round Reversals Outcomes Rate (all) Rate (multi-round) 46 6.4 14 32 2 19 12.4 17.8
Table 2 provides a similar set of measures for the outcomes of the 46 contested City Council primary elections. Those elections witnessed an average of over six candidates, but over a quarter of them were resolved in the first round. But of the 32 elections that required RCV tabulations beyond the first round, only two cases generated a winner different from the candidate who secured a plurality in the first round. One of these was in District 9 where Kristin Richard Jordan squeaked by incumbent Bill Perkins by 114 votes after 13 rounds of tabulation. But Jordan ultimately only received support from 32% of the total ballots cast in the election – a considerably lower percentage than the 36% of ballots that were exhausted! And District 9 was not alone in producing winners who failed to receive a majority of the votes cast and high rates of ballot exhaustion. In fact a majority – 19 elections – of the multi-round RCV calculations resulted in non-majority winners. Figure 1 provides the distribution of the winner’s share in RCV-tabulated races and shows that there were a large number of outcomes in which the winner failed to get as high as even 40% of the total ballots cast.

Of course, the reason that such a large share of races lacked an overall majority winner was that exhaustion rates in the council elections were extraordinarily high. Table 2 reports two measures. The first is the average exhaustion rate for all contested council elections,

12.4%. But of course, races that were resolved in the first round have no exhausted ballots.

So if we focus on the cases that required RCV tabulation, the rate is much higher at

17.8%.13
One objection to my focus on non-majority outcomes and wasted votes under RCV is 13As I discuss below, there is a way to measure “potential exhaustion” in single round outcomes by considering whether a voter’s ballot would have been exhausted had the majority threshold not been met in the first round. Such calculations are not included in Table 2. 11 0 1 2
3 4
5 Frequency
30 40 50 60 70
Winner's Percentage of Votes Cast
Figure 1: Winner’s Share of Votes Cast Figure includes those races which required multiple rounds of tabulation.
0 1 2
3 4 Number of Districts
0 10 20 30 40
Exhaustion Rate
Figure 2: Ballot Exhaustion Rates in NYC Council Democratic Primary Races Figure includes those races which required multiple rounds of tabulation. 12 that similar problems clearly exist with plurality voting. Plurality voting generates nonmajority outcomes and suffers from the problem of “wasted votes” – those going to a candidate with little chance of winning. To address this concern, I collected data on the 33 contested Democratic City Council primary elections in 2017. From that data, I can compute the vote shares of the winner and the number of wasted votes (which I define as a vote for a candidate other than the winner and the runner-up).

As it turns out, the data confirm that rates of vote exhaustion under RCV were approximately the same as the average wasted votes under plurality. Focusing on all contested races in both elections, the average percentage of exhaustion across districts was 12.4% while the average wasted votes was 12.2%. Moreover, the average winning percentage under plurality in 2017 was higher than the percentage under RCV (58.8% versus 52.3%).

Similarly, plurality voting produced a lower rate of non-majority winners than RCV (33.3% versus 41.3%).

So why did RCV fail to outperform plurality voting on these key metrics? The answer is that RCV generated high levels of candidate entry. In 2021 under RCV, 46 nominations were contested with an average of 6.4 candidates per race. In 2017, only 33 districts were contested with 3.4 candidates per race. In the appendix, I report on an analysis of differences between RCV and plurality in terms of majority support and the differences between exhaustion and “wasted votes” as a function of the number of candidates. The analysis of winner’s share demonstrates that those shares fall as the number of candidates rise under both RCV and plurality. Understandably, the ability to transfer votes under RCV reduces the impact of additional candidates compared to plurality.14 However, there is no direct impact of RCV, and the winners’ shares for races with 2 to 5 candidates do not vary systematically across systems. RCV only outperforms when there are a very large number of candidates. But in those cases RCV is never sufficient to produce a majority 14See Figure 6 and Table 8.
13 winner.

The results on exhaustion and wasted votes provided in the appendix are similar.15 There is no statistically significant direct association with RCV, but the system does reduce the impact of additional candidates. Under plurality, each additional candidate is associated with a six percentage point increase in the number of wasted votes while such an increase was three points under RCV. So it appears that the dramatic proliferation of candidates under RCV led exhaustion rates in 2021 to be higher than wasted votes in 2017.

In summary, the comparison on the 2017 and 2021 elections does show that RCV is associated with a larger number of candidates in Democratic primaries for the City Council.

But many of these additional candidates were not very competitive and served primarily to boost ballot exhaustion and non-majority winners. At the same time, RCV tabulations proved pivotal in very few elections – and those involved large numbers of candidates and high exhaustion rates.

3.1 Race, Ethnicity, and Exhaustion
As demonstrated in the last section, the evidence is consistent with the idea that RCV (at least as implemented in New York City) entails a trade-off of more candidates at the cost of more non-majority winners and high rates of ballot exhaustion. Such a trade-off might be deemed acceptable if the negative consequences were born uniformly across all demographic groups. But given previous findings suggesting that racial and ethnic minorities may be more prone to ballot exhaustion, I now examine whether ballot exhaustion rates were higher in electoral districts with higher concentrations of minority voters.16 Ideally, one would like to have individual data on ballot exhaustion linked to the racial, ethnic, and other characteristics of the voter as this would allow us to ask whether certain 15See Figure 5 and Table 8.
16The existence of such racial and ethnic disparities might be grounds for challenging RCV on equal protection grounds (see Bryer (2021)).
14 types of voters are more likely to truncate their ballot in ways that lead to exhaustion.

Unfortunately, although individual cast vote records (CVR) are available, they cannot be linked to individual demographic data for privacy reasons. Thus, scholars are forced to use aggregated data on demographics and make ecological inferences about their relation to vote choice. Of course, ecological inferences face several challenges. Ecological models may be prone to omitted variable bias if neighborhood composition is correlated with unobserved factors that correlate with vote choice. A similar problem may arise to the extent to which the ethnic/racial composition directly effects vote choice. For example, suppose we observed that electoral districts with large Asian populations had large numbers of exhausted ballots. This pattern could arise in two ways – Asian voters could be more likely to cast truncated ballots or other groups who live in proximity to Asians might truncate ballots more frequently. Another common issue is that local demographic information may not distinguish between voters and non-voters.

While these concerns are impossible to address completely, the use of the NYC voter file can ameliorate them to some extent. From information provided by the commercial data company L2, I am able to estimate the aggregate racial and ethnic characteristics of the participants in the 2021 primary at the level of the electoral district. Given that electoral districts are quite small and that there is substantial residential segregation, the ecological biases should be reduced relative to the use of larger and more diverse population units.17 Over the 5633 electoral districts in my data, the average number of 2021 primary voters is 180. There are very few districts with more than 400 (only 118 or 2.1%).18 Of course, the low level of aggregation and the use of the demographics of the actual 17There are slight differences between the data from voter file and the cast vote records which I discuss in the appendix. For example, the raw NYC voter file of 2021 primary voters contains 948,340 records whereas the cast vote record file contains 951,444 votes. At the electoral district level, there are some discrepancies between the voter file totals and the number of cast votes for very small districts. Presumably, these are the consequence of the NYC Board of Elections pooling the tabulation of some smaller districts. The results I present below are unaffected by the exclusion of such districts. 18See Appendix Table 7 for the distribution of primary voters across districts. 15 primary voters does not completely eliminate concerns about ecological biases. But those measures in addition to the robustness of the results across a number of elections and ethnic groups lead some credence to the idea that the findings can be attributed to individual-level behavior.

The models that I report below are based on the following variables:
Ballot Exhaustion Rate. This measure is computed directly from the cast vote record file. After processing the data to account for NYC rules for legal ballots, I determine whether or not a voter cast a ballot ranking one of the two final candidates in a multi-round election.Those who did not cast such a ballot are coded as exhausted and the rate is calculated as the ratio of exhausted ballots to legal ballots in the electoral district. This rate is multiplied by 100 for ease of interpretation. Average Number of Ranks and Rate of Single-Rank Ballots. Using the CVR file, I compute the average number of ranks used by voters in each district along with the proportion of voters who rank only a single candidate. The proportion of single-rank ballots is multiplied by 100 for ease of presentation.
Voter Race and Ethnicity. Data on voter race and ethnicity is provided by L2 and based on a proprietary model that estimates the likelihood of racial and ethnic identities of individual voters from surname and geographical location.19 The available categories are White, African-American, Hispanic (including Portuguese), Asian (East and South), and Other. Within the Other category, L2 identifies several ethnic or nationality groups. The largest are Arab, Russian, Pakistani, Persian, and Filipino.

Although it might make sense to treat the Other category as a combination of ethnic minority groups, I will not emphasize those results given the heterogeneity of the 19In future iterations of this paper, I hope to replace the L2 data with that computed from open source models.
16 classification. Some voters are not classified when the model does not generate a clear prediction. I label these “unknown”.
Candidate Race and Ethnicity. Using web searches, I categorized the racial and ethnic identity all of the RCV finalists. The categories include White, Black, Asian, and Hispanic. In may cases, the candidate stress multiracial identities such as Afro-Latino.

In such cases, I categorized the candidate as both Black and Hispanic. Number of Candidates. This data is generated from the certified election reports available form the NYC Board of Elections website.

The results I present are multiple regression models of exhaustion rates (or average votes or single rank ballots) on the shares of each racial and ethnic group in the primary electorate of each voting district. The base category is the share of White voters so that each coefficient represents the increase or decrease in exhaustion relative to 100% White electorates. In all cases, the models include borough fixed effects and robust standard errors.

Table 3 reports the results of the baseline specification for the city- and borough-wide elections.20. First, consider column 1 which reports the results from the mayor’s race.

Here we find that the concentration of all ethnic groups other than Blacks increased the district exhaustion rate substantially. For example, every ten percentage point increase in the Asian population increased the exhaustion rate by about four points. This is rather large effect given that the average exhaustion rate was only 14 points. A similar ten point increase in Hispanic voters is associated with a 1.5 point increase in the exhaustion rate. The Other electorate is also consistently associated with a substantial increase in the exhaustion rate. The only non-White group associated with lower exhaustion in the 20Because the Pubic Advocate’s race was resolved in the first round, there were technically no exhausted votes. But I include it for comparison purposes and define exhaustion as not voting for one of the top two candidates
17 mayor’s race was African-Americans. But this is largely a function of high-levels of Black support for Eric Adams, the African-American candidate who ultimately won. Clearly, any voter who ranked Adams first could not exhaust her ballot. In the comptroller’s race and the combination of borough presidencies (columns 2 and 3), the percentage of Black voters in an electoral district did correlate strongly with ballot exhaustion. Moreover, the findings about Asian and Hispanic electorates is repeated. The findings for Public Advocate replicate those of Mayor – higher exhaustion rates in Asian and Hispanic areas but no effect of the concentration of Black voters, presumably in part because an AfricanAmerican candidate, Jumaane Williams, was declared the winner in the first round of tabulation.

Table 3: Exhaustion
Mayor Boro Pres Comptroller Advocate
Black -0.040∗∗∗ 0.254∗∗∗ 0.151∗∗∗ 0.001
(0.004) (0.011) (0.008) (0.001)
Asian 0.385∗∗∗ 0.185∗∗∗ 0.446∗∗∗ 0.044∗∗∗
(0.009) (0.012) (0.013) (0.003)
Hisp 0.155∗∗∗ 0.207∗∗∗ 0.296∗∗∗ 0.116∗∗∗
(0.006) (0.013) (0.011) (0.003)
Other 0.235∗∗∗ 0.380∗∗∗ 0.382∗∗∗ 0.087∗∗
(0.022) (0.028) (0.021) (0.028)
Unkown -0.170∗∗∗ 0.270∗∗∗ 0.430∗∗∗ 0.012∗
(0.012) (0.036) (0.033) (0.006)
N 5620.000 5620.000 5620.000 5620.000
R-Squared 0.652 0.572 0.419 0.510
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
As noted above, the conventional measurement of exhaustion is limited in that any ballot with a first-place rank for one of the top two candidates cannot, by definition, be 18 exhausted. Thus, the measure conflates a group’s support for one of the top candidates with its members’ willingness to take full advantage of the RCV ballot. To help isolate these distinct factors, Table 4 repeats the analysis with a measure I call “adjusted” exhaustion.

This measure is a calculation of the exhaustion rate for those voters who do not give the highest ranking to one of the two leading candidates. For example, in the Mayor’s election, it is the percentage of exhausted ballots among those who did not rank Eric Adams or Kathryn Garcia first. Specifically, the adjusted exhaustion measure helps to determine whether the lower exhaustion rate in Black districts reported in Table 3 is a result of higher support for Adams in those districts. The results shown in Table 4 generally confirm this possibility.21 Among voters who did not rank Adams or Garcia first, the exhaustion rates correlate strongly with the concentration of Black voters. The other results mirror those of Table 3 except the comptroller’s race where the adjusted exhaustion rate is lower in Black electoral districts.

To briefly summarize, the results are very consistent with high exhaustion rates in Asian, Hispanic, and Other electoral districts. The results regarding Black electorates are contingent on whether there is a Black candidate among the top two vote-getters. In races such as the mayoral and advocate contests with Black finalists, Black districts exhibited slightly lower exhaustion rates than White districts. But in other races, where Black candidates were not finalists, Black districts exhibited higher rates of ballot exhaustion.

The appendix reports on three additional analyses designed to explore the underlying behavior behind the higher exhaustion rates in heavily minority primary electorates. Table 9 repeats the analysis but focuses instead on the number of ranks used by voters in primary electorates of different compositions. I find that heavy concentrations of Asian and Hispanic voters reduces the average number of ranks used by voters relative to White districts in all elections except for the lopsided Advocate’s race. The pattern for Black 21I do not include the Advocate’s race because the adjusted exhaustion rate is very high in every district. 19 Table 4: Adjusted Exhaustion
Mayor Boro Pres Comptroller
Black 0.047∗∗∗ 0.133∗∗∗ -0.042∗∗∗
(0.006) (0.010) (0.009)
Asian 0.403∗∗∗ 0.064∗∗∗ 0.183∗∗∗
(0.010) (0.015) (0.014)
Hisp 0.260∗∗∗ 0.126∗∗∗ 0.059∗∗∗
(0.009) (0.012) (0.012)
Other 0.366∗∗∗ 0.311∗∗∗ 0.222∗∗∗
(0.022) (0.027) (0.018)
Unkown -0.173∗∗∗ 0.003 -0.021
(0.018) (0.033) (0.038)
N 5608.000 5597.000 5605.000
R-Squared 0.525 0.587 0.215
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
electorates is somewhat different – fewer ranks in the Mayor’s race and slightly more in the others. While one would not want to overinterpret these results, this pattern may be consistent with English-language proficiency playing an important role.

A second analysis reported in Table 10 looks at the proportion of voters in each district that cast a ballot ranking only a single candidate. Here the results are also mixed. In the Mayor’s race, concentrations of racial and ethnic voters are associated with a strong increase in the number of voters choosing only a single candidate. However, this is not true about Black electorates in the other races where Black electorates voted for a single candidates at rates similar to those of White electorates (again the lopsided Advocate’s race stands out as an anomaly). Table 11 examines the effect of the NYC’s limit of five candidate rankings per ballot. For the elections that required multiple rounds of tabulation, I estimate two models. The first is based on those voters who fully participated in the election by casting 20 the minimum of 5 or n − 1 ranks in a n candidate election.22 The second model is based on the exhaustion of those who truncated their ballot to a degree not required by the ballot design (less than the minimum of 5 or n − 1). The correlations of the minority voter percentages and exhaustion rates are statistically significant and have the same sign for both groups of voters. However, the magnitudes are far larger for the voters who voluntarily truncated by ranking fewer than five candidates. These results suggest that NYC’s RCV primary would have had substantial racial and ethnic disparities in exhaustion even if the ballot had allowed voters to rank more candidates.

3.2 City Council Elections
In this section, I extend the analysis of the correlates of exhaustion to the 32 City Council primaries that were resolved though multiple-round tabulations. This analysis allows me to better control for additional factors such as the total number of candidates in a council district as well as the racial and ethnic identities of the finalists. Table 5 reports these results. Column 1 uses a similar specification as the one for the executive elections, but controls for the number of candidates. Not surprisingly, given the discussion above, the number of candidates strongly correlates with ballot exhaustion. Each individual candidate is associated with almost a two percentage point increase in the exhaustion rate. But more importantly, districts with large racial and ethnic minority populations have considerably higher rates of ballot exhaustion.

The second column conditions on the race and ethnicity of the final two candidates.

The model includes both the unconditional effect of the candidate’s identity as well as an interaction between the candidate’s identity and the concentration of co-racial/ethnic voters in each district. The unconditional effect allows exhaustion to go up or down in each district independent of its composition, and the interaction effect allows for the possibility 22Voting for n − 1 of n ensures that one’s ballot cannot be exhausted. 21 that voters are less likely to exhaust if there is a co-ethnic finalist.

The unconditional effects show that the identity of the finalists has a substantial impact on exhaustion rates. The presence of a Black finalist is associated with an exhaustion rate five points greater than a contest with two White candidates. Asian candidates are associated with a boost of 2.6 points while Hispanic candidates coincide with a 3 point reduction. All three interaction effects are negative and statistically significant. This finding is consistent with the idea that voters are less likely to exhaust when there is a co-ethnic/racial candidate in the final round.

When I combine the coefficients on the percentage of each group with the associated interaction effect, the effect of percentage Black on ballot exhaustion in the presence of a Black finalist remains negative but is not statistically significant. However, the effect of percent Asian and Hispanic are statistically significant at the .1% in the presence of a co-ethnic finalist.23 The coefficients for the multi-language groups are larger than that of Blacks, suggesting that ballot complexity may be a contributing factor.

To illustrate the magnitudes of these relationships, Figure 3 presents the predicted differences in the exhaustion rates between hypothetical homogeneous minority districts and those of 100% White districts. In the cases where there is no co-ethnic finalist, the figure shows that homogeneous Black, Asian, and Hispanic districts would have 12 to 20 points higher exhaustion rates than a homogeneous White district (holding the number of candidates and the borough constant). When there are co-ethnic finalists, the exhaustion rates in homogeneous Asian and Hispanic districts remain 6 to 8 percentage points higher, but the rates in Black districts are indistinguishable from White districts.

Column three presents a different model where I condition the effects of racial composition on the number of candidates. These interactions of the district composition for all three groups are broadly similar. For the minimum number of candidate in RCV tabulated 23The estimated condition effect for Percent Asian is .067 with a standard error of .017. For Percent Hispanic, it is .084 with a standard error of .018.
22 Table 5: Exhaustion in Council Primaries
Baseline Candidate Demographics Number of Candidates
# of Candidates 1.924∗∗∗ 1.865∗∗∗ 0.984∗∗∗
(0.056) (0.054) (0.152)
% Black 0.035∗∗ 0.174∗∗∗ -0.027
(0.012) (0.018) (0.018)
% Asian 0.080∗∗∗ 0.190∗∗∗ -0.165∗∗∗
(0.016) (0.031) (0.046)
% Hispanic 0.076∗∗∗ 0.131∗∗∗ -0.154∗∗∗
(0.016) (0.022) (0.040)
% Other 0.371∗∗∗ 0.336∗∗∗ 0.359∗∗∗
(0.051) (0.055) (0.051)
% Unknown 0.061 0.091 0.050
(0.055) (0.055) (0.054)
Black Cand in Final Rnd 5.304∗∗∗
(0.657)
Black Cand x % Black -0.185∗∗∗
(0.016)
Asian Cand in Final Rnd 1.672∗∗
(0.611)
Asian Cand x % Asian -0.122∗∗∗
(0.031)
Hispanic Cand in Final Rnd -2.990∗∗∗
(0.637)
Hispanic Cand x % Hispanic -0.047∗
(0.022)
# of Candidates × % Black 0.008∗∗∗
(0.002)
# of Candidates × % Asian 0.033∗∗∗
(0.007)
# of Candidates × % Hispanic 0.029∗∗∗
(0.005)
N 3552.000 3552.000 3552.000
R-Squared 0.376 0.429 0.387
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
23 100% Black (No Co-ethnic)
100% Asian (No Co-ethnic)
100% Hispanic (No Co-ethnic)
100% Black (Co-ethnic)
100% Asian (Co-ethnic)
100% Hispanic (Co-ethnic)
-10 0 10 20 30
Differences with 100% White District
No Co-ethnic Finalist Co-ethnic Finalist
Figure 3: Predicted Exhaustion Rate Differences with 100% White District Figure includes those races which required multiple rounds of tabulation. elections (3), there is very little difference across districts based on ethnic composition. But as the number of candidates gets larger, there is substantial variation related to composition. Figure 4 illustrates the magnitudes of the racial and ethnic composition coefficients as a function of the number of candidates appearing on the ballot. The effect of the number of candidates on the magnitude of difference between 100% White districts is especially large for Asian and Hispanic districts. Such a pattern is consistent with minority voters being disadvantaged by the candidate proliferation of RCV much more than White voters.

The data on City Council races allows me to test an additional claim about RCV – that it boosts voter engagement. If this claim were true, we would expect to see less voter roll-off between the higher-profile executive races and the down-ballot council races.

However, since previous research has shown that ballot completion is less likely when voters have less information and sophistication (Lamb and Perry 2020), one might expect RCV to exacerbate roll-off given the increased informational demands required to rank multiple 24 -.2
0 .2 .4 Coefficient on % Black
3 6 9 12 15
Number of Candidates -.2
0 .2 .4 Coefficient on % Asian
3 6 9 12 15
Number of Candidates -.2
0 .2 .4 Coefficient on % Hispanic
3 6 9 12 15
Number of Candidates
Figure 4: Coefficients on Racial and Ethnic Composition and % of Candidates candidates.24
The data are much more consistent with the latter perspective as we observe very high roll-off rates in the council races. I find that 9.3% of the voters who voted in the mayor’s race failed to cast a vote in one of the 32 RCV tabulated council races. This is much higher than the 5.7% of 2017 voters who skipped voting in a contested council primary.25 This gap is consistent with the voters being deterred by the use of RCV in the lower profile elections. But for my purposes, it is important to know whether rates of roll-off 24The question of roll-off is distinct from whether RCV induces more voters to show up at the polls in the first place. McDaniel (2016) finds that turnout dropped in San Francisco mayoral elections following the adoption of RCV, especially among minority groups. In a study of several RCV cities matched against comparable plurality cities, Kimball and Anthony (2016) find a four percentage point drop in turnout associated with RCV, although the estimate is not statistically significant on its own. In a more recent study, McDaniel (2019) finds a statistically significant five percentage point drop due to the introduction of RCV in municipal elections relative to similar cities that maintain plurality electoral systems. While there is disagreement about the magnitude and statistical reliability of the estimated declines in voter turnout, I am not aware of any study that finds a boost in turnout associated with switching to RCV from plurality voting.
25One might object that that these numbers are not strictly comparable because some voters may have skipped voting in the 2017 mayor’s election and then voted in a council race. But given the low number of voters skipping the mayor’s race, such voters would have a very small effect on the comparison. 25 varied significantly across electoral districts based on their racial and ethnic composition.

So Table 12 in the Appendix reports a model of ballot roll-off similar to those reported above. The results show that ballot roll-off rates were higher in electoral districts with larger Black, Asian, and Hispanic populations. As with exhaustion rates, roll-off rates are attenuated in the presence of co-ethnic finalists. The conditional effects remain positive for Black and Hispanic populations, but roll-off rates in Asian districts are slightly smaller when there is a competitive Asian candidate on the ballot.26 In summary, the evidence shows that ballot exhaustion was endemic in the NYC Council primaries. Rates were especially high in elections with large numbers of candidates despite the fact that the vast majority of the incremental candidates were not at all competitive.

The findings are also consistent with large gaps in the ballot exhaustion rates across ethnic and racial groups, especially groups with large numbers of non-native English speakers. The gaps were partially attenuated when co-ethnic candidates were strong enough to advance to the final round. Given Colner (2023)’s finding that the boost to the diversity pool associated with the adoption of RCV is modest and short-lived, the findings about coethnic candidates may portend greater problems for RCV in the future. 4 Alaska Top Four Primary and RCV General
To explore the generality of the findings from New York City, I now examine a very different context – the 2022 Alaska state and federal elections. In 2020, the voters of Alaska adopted Measure 2 which replaced their existing electoral system of first-past-the-post (FPTP) 26These observational results can be usefully contrasted with recent experimental results. In Ntounias (2023), respondents are assigned to vote in a hypothetical RCV election while a control group votes in a plurality election with the same candidates. He finds that RCV voters are no more likely to search for candidate information (i.e. read candidate bios) and are equally likely to abstain. Thus, the opportunity to rank more candidates had no important impact on voter engagement. Experimental evidence from international settings reported by Blais et al. (2021) also confirms that ranked choice does not impact voter satisfaction.
26 partisan primaries followed by FPTP general elections with a two-stage process where the top four candidates from a “pick one” open primary compete in a general election conducted under RCV rules. The measure passed narrowly with a margin of less than 4000 votes of the roughly 344,000 valid ballots cast.27
Following the passage of Measure 2, Alaska implemented its Top Four system for the first time in 2022. The primary stage of these elections was almost completely inconsequential.

While the top races did involve large numbers of candidates (US Senate 19, US Rep 22, and Governor 10), only one of the 59 state legislative primaries had more than four candidates (HD 35 which had 5). Even with the large number of candidates, the top of the ticket primaries were almost as meaningless. The fourth place candidates polled 2.13% in the Senate race, 3.77% in House race, and 3.86% in the governor’s race. On average, they trailed the leading candidates by over 35 points, but only beat the fifth place candidate by about two points. In the general election, the fourth candidates tallied a mere 3% of the vote in the first round. Given the trivial outcomes of the primaries, I focus on the role of RCV and ballot exhaustion in the general elections.

Table 6 reports the rates of ballot exhaustion in the 2022 Alaska elections. The table reports exhaustion rates both for elections that involved RCV calculations as well as “potential rates” for elections that were resolved in the first round. Across all elections, the data show that around 1 of 20 Alaskan voters who went to the polls had no vote recorded in the last round of tabulation. While these rates are somewhat lower than those in NYC, it is worth noting that there were many fewer candidates in Alaska’s Top 4 setting. But of course, it is clear that the four candidate limitation did not eliminate high rates of exhaustion. Moreover, the rates of exhaustion in Alaska generally exceed those in similar RCV elections in Maine (see McCarty (2020).)
27Brooks, James (November 18, 2020). “Alaska becomes second state to approve ranked-choice voting as Ballot Measure 2 passes by 1%”. Anchorage Daily News. Retrieved December 6, 2022. Hillman, Anne (September 15, 2022) “Why Alaska uses ranked choice voting and what we know about how it affects elections”. Alaska Public Radio, retrieved March 6, 2023. 27 Table 6: Exhausted Ballots in Alaska Elections
Election % Exhausted
House Special 6.0%
House General 5.5%
Senate 3.4%
Governor (potential) 6.1%
State Senate 4.9%
State Senate (potential) 7.5%
State House 3.7%
State House (potential) 6.5%
State House (four candidates) 4.8%
Potential exhaustion rates include those voters whose ballots would have been exhausted had the election proceeded to RCV tabulation As I discussed in the context of New York City, high exhaustion rates might arguably be acceptable if voters from across all racial, ethnic, and social groups were roughly equally likely to exhaust their ballots. So the main question is whether or not certain groups of Alaskans are more or less likely to exhaust their ballots. Therefore, as I did for NYC, I conduct a multivariate regression analysis of the correlates of ballot exhaustion. The primary dependent variable is the ballot exhaustion rate at the precinct level. Using CVR data provided by the State of Alaska Division of Elections, I was able to determine whether each ballot in the cast vote record file was exhausted in a given race.28 I then aggregate these data to the precinct level. I use two versions of this dependent variable. The first, labeled “raw,” is the total rate of exhausted ballots in a precinct for a given election. The problem with this measure, however, is that it might be conflated with partisanship as a voter who ranks one of the final two candidates (generally a major party candidate) first cannot exhaust her ballot regardless of how many candidates she chooses. So as I did for the NYC executive elections, I use a second “adjusted” measure which is the rate of exhaustion among those voters who do not rank one of the final two candidates first. This measure has the advantage of eliminating the effects of ranking a top candidate 28See https://www.elections.alaska.gov/election-results/e/?id=22genr. 28 first, but it is noisier because it is based on a much smaller number of voters. Below I report both sets of results. Another issue is that I can only calculate exhaustion rates for elections that use RCV tabulations. Therefore, I used “potential” exhaustion in the governor’s race and state legislative districts with more than two candidates. This measure is based on a determination of whether the voter’s ballot would have been exhausted had the leading candidate failed to reach 50%. Precincts in state legislative districts with only two candidates cannot be included in the analysis.

Unfortunately, the L2 data file does not contain reliable data on race and ethnicity in Alaska, so I must aggregate census block and block group data to the precinct level.29 From the Census data, I compile racial and ethnic shares of Whites, Blacks, Asians, Hispanics, Alaskan Natives, and Other.30 The use of Census data has limitations. The level of aggregation is higher, and the measures are based on the full population rather than voters.

But the estimated relationships may still be informative especially in light of the findings with higher quality data in NYC. To help offset some of these concerns, I include additional demographic data on education and age. I include education variables for less than high school degree, high school degree, some college, and college graduate. I also calculate the share of precinct residents over the age of 65.

Finally, it is worth noting that Alaska presents a fairly difficult case for detecting racial and ethnic differences in ballot exhaustion. Since the Top 4 system serves to limit the number of candidates on the RCV ballot, Alaska is unlikely to produce the large cross-group variations we saw in NYC with its larger number of candidates. Yet the results presented below do provide compelling evidence of ethnic group differences even in elections with a lower number of candidates.
29This problem largely stems from the historical surname adoption practices of Alaskan Natives. See https://namecensus.com/last-names/common-american-indian-and-alaskan-native-surnames/. 30The data on Alaskan Natives is based on the Census classification American Indian/Alaskan Native.

For simplicity, I refer to the group simply as Alaskan Natives, conceding that a small share of this group may consist of members of non-Alaskan tribal groups.
29 My main analyses are regressions of the exhaustion rates on the racial, educational, and age variables. In the models of state house elections, I also include the number of candidates in the election (this is a constant in all of the other cases including state senate elections, none of which had more than three candidates). My analysis did not uncover any systematic correlations between education levels and racial demographics other than Alaskan Native. However, I found that the percentage of Alaskan Natives was consistently highly correlated with exhaustion rates across electoral contexts. Tables 13–18, found in the appendix, present these findings in full.

Table 7 presents the full set of coefficients on the share of Alaskan Natives in a precinct across all of the elections. Panel 1 reports the estimates for the special US House election held in July of 2022. In this election, Mary Peltola, an Alaskan Native, won a four-way race against three Republican candidates including former Vice-Presidential candidate Sarah Palin and Nick Begich III, a member of a prominent political family. Not surprisingly, given Peltola’s co-ethnicity, there is no correlation between the Alaskan Native share and the exhaustion rate. However, the exhaustion rates of those ranking Begich first were high and somewhat concentrated in Alaskan Native precincts, although the coefficient is only significant at the p < .10 level. One might argue that the success of an Alaskan Native candidate rebuts the idea that RCV is bad for minority-group voters. But the pattern of voting on the RCV ballots indicate that the outcome would likely have been the same under the prior system of a partisan primary followed by a general election.31 Panel 2 reports the findings from the House general election in November of 2022.

This election was generally a replay of the special election.32. The pattern of exhaustion is just a clearer version of what happened in the special election. In the general election, Alaskan Native support for Peltola was strong enough to reduce the exhaustion rate for their 31Assuming RCV rankings are indicative of how voters would have voted in primary and general elections, Palin would have clearly received the GOP nomination and lost to Peltola head-to-head in the general. 32The main difference was that the fourth place candidate dropped out of the race, but still appeared on the ballot
30 Table 7: Coefficients on Alaska Native Population
Exhaustion Exhaustion Adj
Special House Election 0.007 0.052
(0.009) (0.027)
N 369 369
R2 0.068 0.116
House General Election -0.031∗∗∗ 0.205∗∗∗
(0.008) (0.029)
N 374 370
R2 0.186 0.312
Governor 0.125∗∗∗ 0.181∗∗∗
(0.012) (0.028)
N 374 372
R2 0.435 0.409
US Senate 0.013∗ 0.033
(0.005) (0.033)
N 374 369
R2 0.128 0.156
State Senate 0.076∗ 0.156
(0.036) (0.092)
N 141 141
R2 0.139 0.150
State House 0.258 1.117∗
(0.168) (0.518)
N 91 91
R2 0.316 0.288
Models include shares of other racial and ethnic groups, senior citizens and educational attainment groups.

Potential exhaustion rates used for Governor and state legislatures.

State house model controls for number of candidates.
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
31 precincts, but the exhaustion rate for those who ranked Begich first are heavily concentrated in Alaskan Native areas. The correlation is highly significant.

Panel 3 reports on the Governor’s race and is based on potential exhaustion since the winner, Michael Dunleavy obtained 50.2% of the first ranks. But had Dunleavy received about 800 fewer voters, 6% of the ballots would have been exhausted (see above). The regression models show again that these exhausted ballots would have again been concentrated in Alaskan Native precincts.

The US Senate race is reported in Panel 4. Despite the fact that this race had the lowest overall exhaustion rate of any of the elections, I still estimate a small but statistically significant relationship between raw exhaustion and the share of Alaskan Native residents.33 Panels 5 and 6 mostly replicate the above patterns in State Senate and House elections.

Because there were many uncontested and two candidate state legislative races, the models of exhaustion could be estimated using only a fraction of the precincts. To ensure an adequate sample size, I use the potential exhaustion measure for races with three or more candidates that resolved in the first round. Although this loss of statistical power led to a loss of precision in the estimates, the point estimates are substantial and in line with the findings for the other elections.

In summary, despite the fact that the Top Four limited the proliferation of candidates, the results from Alaska are very consistent with the findings from NYC. Across races, there is a statistically significant relationship between ballot exhaustion and the share of the population belonging to a politically-salient minority. Also similar to NYC, the outcomes of the US House election suggest that these relationships may be mitigated when there are strong co-ethnic candidates.
33The coefficient on Alaskan Native is not significant in the adjusted exhaustion model largely due to the fact that just slightly more than 10% did not vote for one of the top two candidates. Thus, the adjusted measure is very noisy in this case.
32 5 Conclusion
Partisan polarization and related political dysfunctions have greatly increased interest in fundamental reforms to our electoral processes. But such reforms often come with a variety of trade-offs and unintended consequences. So careful scrutiny and evaluation of the effects of reforms is crucial.

Ranked choice voting is a reform that has excited a lot of people. Its advocates suggest that it can both turn down the temperature surrounding electoral politics and increase the diversity of choices available to voters. But scholarly attempts to evaluate such claims and to root out any downsides are still in their infancy. In this paper, I try to evaluate one such downside – the high rates of ballot exhaustion and their concentration in precincts with large minority populations. The findings suggest that these are indeed drawbacks of RCV. Across a variety of electoral contexts in New York City and Alaska, I find consistent correlations between the ethnic and racial composition of a precinct and the share of exhausted ballots. These correlations are especially large when there are large numbers of candidates and when there are not strong co-ethnic candidates in the race.

These findings are consistent with RCV providing an advantage to majority-group voters over minority-group voters. Whereas RCV allows majority-group voters an additional opportunity to resolve candidate coordination problems, the patterns of ballot exhaustion suggest that minority-group voters are not taking full advantage the system. Whether those higher rates of exhaustion are due to ballot complexity, lower levels of information and mobilization, or racial and ethnic polarization, it is clear that the potential effects of RCV on minority voters needs to be carefully scrutinized before further adoption. 33 References
Blais, Andr´e, Carolina Plescia, John H¨ogstr¨om and Gabrielle P´eloquin-Skulski. 2021. “Do (Many) Voters Like Ranking?” Party Politics 27(6):1223–1228.

Bryer, Brandon. 2021. “One Vote, Two Votes, Three Votes, Four: How Ranked Choice Voting Burdens Voting Rights and More.” University of Cincinnati Law Review 90:711.

Buisseret, Peter and Carlo Prato. 2022. “Politics Transformed? How Ranked Choice Voting Shapes Electoral Strategies.” Unpublished manuscript.

Burnett, Craig M. and Vladimir Kogan. 2015. “Ballot (and Voter) “Exhaustion” under Instant Runoff Voting: An Examination of Four Ranked-Choice Elections.” Electoral Studies 37:41–49.

Coll, Joseph. 2021. “Demographic Disparities Using Ranked-Choice Voting? Ranking Difficulty, Under-Voting, and the 2020 Democratic Primary.” Politics and Governance 9(2):293–305.

URL: https://www.cogitatiopress.com/politicsandgovernance/article/view/3913 Colner, Jonathan. 2023. “Running Towards Rankings: Ranked Choice Voting’s Impact on Candidate Entry and Descriptive Representation.” Available at SSRN 4317298 .

Cormack, Lindsey. 2023. “More Choices, More Problems? Ranked Choice Voting Errors in New York City.” American Politics Research OnLine First.

URL: https://doi.org/10.1177/1532673X231220640
Drutman, Lee. 2020. Breaking the Two-Party Doom Loop: The Case for Multiparty Democracy in America. Oxford University Press, USA.

Kilgour, D. Marc, Jean-Charles Gr´egoire and Ang`ele M. Foley. 2020. “The Prevalence and 34 Consequences of Ballot Truncation in Ranked-Choice Elections.” Public Choice 184:197– 218.

Kimball, David C. and Joseph Anthony. 2016. Voter Participation with Ranked Choice Voting in the United States. In Annual Meeting of the American Political Science Association.

Lamb, Matt and Steven Perry. 2020. “Knowing What You Don’t Know: The Role of Information and Sophistication in Ballot Completion.” Social Science Quarterly 101(3):1132– 1149.

McCarty, Nolan. 2019. Polarization: What Everyone Needs to Know. New York: Oxford University Press.

McCarty, Nolan. 2020. “Expert Report, Hagopian, et al. v. Dunlap, et al.” U.S. District Court for the District of Maine .

McDaniel, Jason A. 2016. “Writing the Rules to Rank the Candidates: Examining the Impact of Instant-runoff Voting on Racial Group Turnout in San Francisco Mayoral Elections.” Journal of Urban Affairs 38(3):387–408.

McDaniel, Jason A. 2018. “Does More Choice Lead to Reduced Racially Polarized Voting? Assessing the Impact of Ranked-Choice Voting in Mayoral Elections.” California Journal of Politics and Policy 10(2).

McDaniel, Jason A. 2019. “Electoral Rules and Voter Turnout in Mayoral Elections: An Analysis of Ranked-Choice Voting.”.

Morton, Rebecca B. and Thomas A. Rietz. 2007. “Majority Requirements and Minority Representation.” NYU Annual Survey of American Law 63:691. 35 Neely, Francis and Corey Cook. 2008. “Whose Votes Count?: Undervotes, Overvotes, and Ranking in San Francisco’s Instant-Runoff Elections.” American Politics Research 36(4):530–554.

URL: https://doi.org/10.1177/1532673X08318110
Neely, Francis and Jason McDaniel. 2015. “Overvoting and the Equality of Voice under Instant-runoff Voting in San Francisco.” California Journal of Politics and Policy 7(4).

Ntounias, Theodoros. 2023. “Voter Information Search and Ranked Choice Voting.” Election Law Journal: Rules, Politics, and Policy 22(4):337–362.

URL: https://doi.org/10.1089/elj.2022.0052
Polubinksi, Madeleine. 2023. “One Person, How Many Votes? Race and Representation in Ranked Choice Elections.”. Senior Thesis, Politics Department, Princeton University.

Santucci, Jack. 2021. “Variants of Ranked-Choice Voting from a Strategic Perspective.” Politics and Governance 9(2):344–353.

Santucci, Jack and Jamil Scott. 2021. “Do Ranked Ballots Stimulate Candidate Entry?” Available at SSRN 3956554 .

Vishwanath, Arjun. 2022. “Electoral Institutions and Substantive Representation in Local Politics: The Effects of Ranked Choice Voting.”.
36 6 Appendix

6.1 Comparisons of 2017 and 2021
Column 1 of Table 8 reports regression models of the number of exhausted or wasted voters in contested city council primaries in 2017 and 2021. Note that while the direct effect of RCV elections in 2021 is not statistically different from zero, the correlation between exhausted and wasted votes with the number of candidates is lower under RCV.

Column 2 of Table 8 reports regression models of the size of the winning majority in contested city council primaries in 2017 and 2021. Here the models suggest that RCV was associated with substantially lower winners’ shares, but that RCV may have mitigated the effect of the number of candidates.

Table 8: Comparison of RCV and Plurality
Exhaust/Waste Winner’s Share
Candidates 6.427∗∗∗ -5.334∗∗∗
(0.658) (0.921)
RCV (2021) 3.040 -12.119∗
(2.639) (4.899)
RCV x Candidates -3.420∗∗∗ 3.375∗∗
(0.698) (1.003)
Constant -9.851∗∗∗ 77.026∗∗∗
(2.152) (3.951)
N 79.000 79.000
R-Squared 0.752 0.475
Standard errors in parentheses
Regression model
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
Figures 5 and 6 present these results graphically.
37 0 20 40 60
0 5 10 15
Candidates
Figure 5: Exhaustion and Waste Rates by Number of Candidates Diamonds and solid line represent RCV races in 2021 while circles and dotted line represent plurality contests in 2017. 20 40 60 80 100
0 5 10 15
Candidates
Figure 6: Winner’s Share by Number of Candidates Diamonds and solid line represent RCV races in 2021 while circles and dotted line represent plurality contests in 2017. 38 0 200 400 600 800 Frequency
0 200 400 600 800 1000
Number of Primary Voters
Figure 7: Number of Primary Voters by District From NYC Voter File

6.2 Cast Vote Records and the Voter File
Figure 7 shows a histogram of the number of cast vote records for each NYC precinct. Only a handful of precincts had more than 400 voters with most ranging from 100 to 200 voters.

Figure 8 plots the number of records in the CVR with the number of primary voters as recorded in the L2 voter file most proximate to the election. In most cases, the figures closely match. There are some discrepancies which might be attributed to moving or attrition of voters from the voter file soon after the election and/or the NYC Board of Elections combining precincts for tabulation purposes.

6.3 Correlates of the Number of Ranks and the Single Ranks Table 9 re-estimates the models described in the main text by using the number of ranks used by each voter to construct the dependent variable. All observations are weighted by the number of votes cast in each electoral district. The model contains borough fixed effects, and robust standard errors are reported.
39 0 200 400 600 800 1000 Voter File
0 200 400 600 800 1000
Cast Vote Record File
Figure 8: Number of Records in Voter File versus Cast Vote Records with 45 degree line
Table 9: Number of Votes
Mayor Boro Pres Comptroller Advocate
Black -0.008∗∗∗ 0.001∗∗∗ 0.002∗∗∗ 0.002∗∗∗
(0.000) (0.000) (0.000) (0.000)
Asian -0.016∗∗∗ -0.001∗∗∗ -0.002∗∗∗ 0.003∗∗∗
(0.000) (0.000) (0.000) (0.000)
Hisp -0.014∗∗∗ 0.000 -0.001∗∗∗ 0.003∗∗∗
(0.000) (0.000) (0.000) (0.000)
Other -0.010∗∗∗ 0.002∗∗ 0.004∗∗∗ 0.004∗∗∗
(0.001) (0.001) (0.001) (0.000)
Unkown -0.022∗∗∗ -0.009∗∗∗ -0.009∗∗∗ -0.002∗∗∗
(0.001) (0.001) (0.001) (0.000)
N 5620.000 5614.000 5619.000 5616.000
R-Squared 0.399 0.757 0.359 0.522
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
40 Table 10 re-estimates the models described in the main text but uses the number of voters who ranked only a single candidate to construct the dependent variable. All observations are weighted by the number of votes cast in each electoral district. The model contains borough fixed effects, and robust standard errors are reported.

Table 10: Single Vote Ballots
Mayor Boro Pres Comptroller Advocate
Black 0.202∗∗∗ 0.006 -0.010 -0.234∗∗∗
(0.006) (0.006) (0.006) (0.007)
Asian 0.319∗∗∗ 0.021 0.061∗∗∗ -0.396∗∗∗
(0.011) (0.011) (0.011) (0.012)
Hisp 0.323∗∗∗ 0.030∗∗∗ 0.082∗∗∗ -0.349∗∗∗
(0.008) (0.009) (0.008) (0.010)
Other 0.270∗∗∗ -0.018 0.042∗∗ -0.200∗∗∗
(0.017) (0.014) (0.014) (0.017)
Unkown 0.181∗∗∗ -0.106∗∗∗ -0.052∗-0.640∗∗∗
(0.021) (0.021) (0.024) (0.032)
N 5620.000 5620.000 5620.000 5620.000
R-Squared 0.543 0.294 0.153 0.434
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
Table 11 reports estimates of separate models distinguishing between the rates at which voters who fully participated exhausted their ballots (F) and the rate at which those who truncated (T) exhausted their ballots.34 For Asian, Hispanic, and Other electorates, the magnitudes of the coefficients are much higher for voters who voluntarily truncated by not ranking as many candidate as there were allowed. Yet I still find significant correlations between the size of those electorates and exhaustion among those who fully participated in the election.
34The number of electoral districts varies due to districts where every voter cast a full ballot or no voter cast a full ballot.
41 Table 11: By Source of Truncation
Mayor F Mayor T BP F BP2 T Comp F Comp T
Black -0.031∗∗∗ -0.066∗∗∗ 0.091∗∗∗ 0.325∗∗∗ -0.026∗∗∗ -0.028∗∗∗ (0.004) (0.006) (0.004) (0.016) (0.005) (0.005)
Asian 0.081∗∗∗ 0.518∗∗∗ 0.028∗∗∗ 0.288∗∗∗ 0.347∗∗∗ 0.425∗∗∗ (0.007) (0.011) (0.005) (0.022) (0.012) (0.011)
Hisp 0.049∗∗∗ 0.215∗∗∗ 0.032∗∗∗ 0.345∗∗∗ 0.112∗∗∗ 0.207∗∗∗ (0.005) (0.009) (0.005) (0.019) (0.007) (0.008)
Other 0.080∗∗∗ 0.351∗∗∗ 0.082∗∗∗ 0.580∗∗∗ 0.118∗∗∗ 0.332∗∗∗ (0.014) (0.027) (0.011) (0.042) (0.017) (0.027)
Unkown -0.093∗∗∗ -0.322∗∗∗ -0.011 0.219∗∗∗ -0.065∗∗∗ -0.196∗∗∗ (0.013) (0.017) (0.012) (0.047) (0.015) (0.015)
N 5598.000 5602.000 5605.000 5596.000 5585.000 5610.000 R-Squared 0.221 0.649 0.644 0.429 0.477 0.621
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

6.4 Ballot Roll-Off
Table 12 presents the estimates of a model of ballot roll-off in NYC precincts as a function of the number of city council candidates and the racial and ethnic composition of each electoral district. All observations are weighted by the number of votes cast in each district. The model contains borough fixed effects, and robust standard errors are reported. 42 Table 12: Ballot Roll-Off between Mayor and Council Primaries Baseline Candidate Demographics
# of Candidates -0.310∗∗∗ -0.275∗∗∗
(0.032) (0.030)
% Black 0.019∗∗∗ 0.049∗∗∗
(0.004) (0.008)
% Asian 0.012 0.140∗∗∗
(0.007) (0.015)
% Hispanic 0.039∗∗∗ 0.084∗∗∗
(0.007) (0.010)
% Unknown 0.072∗∗∗ 0.052∗∗∗
(0.015) (0.015)
% Other -0.001 -0.032∗∗
(0.013) (0.010)
Black Cand in Final Rnd -1.229∗∗∗
(0.243)
Black Cand x % Black -0.022∗∗
(0.008)
Asian Cand in Final Rnd 1.337∗∗∗
(0.265)
Asian Cand x % Asian -0.178∗∗∗
(0.015)
Hispanic Cand in Final Rnd -2.376∗∗∗
(0.329)
Hispanic Cand x % Hispanic -0.013
(0.012)
N 3552.000 3552.000
R-Squared 0.168 0.260
Standard errors in parentheses
Regression models with borough fixed effects
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
43 6.5 Alaska Models
Tables 13–18 provide estimates of the full specifications for the models of ballot exhaustion in the 2022 Alaskan elections. In each case, the observations are weighted by the number of voters who cast ballots in the precinct. Each model reports robust standard errors. In some cases, the number of precincts used for the adjusted exhaustion model is lower than the number in the raw exhaustion model. This occurs when there are precincts where all voters ranked one of the top two candidates first.

Table 13: Special House Election
Exhaustion Exhaustion Adj
Black -0.158∗-0.384∗∗
(0.064) (0.116)
Hisp 0.018 -0.130
(0.099) (0.191)
American Indian/Alaska Native 0.007 0.052
(0.009) (0.027)
Asian 0.002 0.032
(0.028) (0.050)
Less than HS 0.015 0.013
(0.029) (0.067)
High School 0.005 0.032
(0.012) (0.025)
Some College 0.035 0.081∗
(0.019) (0.037)
Over 65 0.030 0.100∗
(0.019) (0.047)
N 369.000 369.000
R-Squared 0.068 0.116
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
44 Table 14: General House Election
Exhaustion Exhaustion Adj
Black -0.201∗∗∗ -0.611∗∗∗
(0.060) (0.150)
Hisp -0.070 -0.114
(0.093) (0.232)
American Indian/Alaska Native -0.031∗∗∗ 0.205∗∗∗
(0.008) (0.029)
Asian 5.785∗ 35.266∗∗∗
(2.689) (7.073)
Less than HS 0.046 0.172∗
(0.025) (0.080)
High School 0.010 -0.010
(0.013) (0.037)
Some College 0.048∗∗ 0.071
(0.016) (0.041)
Over 65 0.056∗∗ 0.232∗∗∗
(0.018) (0.045)
N 374.000 370.000
R-Squared 0.186 0.312
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
45 Table 15: Governor Election
Exhaustion Exhaustion Adj
Black -0.322∗∗∗ -0.559∗∗
(0.069) (0.186)
Hisp -0.007 -0.209
(0.092) (0.242)
American Indian/Alaska Native 0.125∗∗∗ 0.181∗∗∗
(0.012) (0.028)
Asian 0.141∗∗∗ 0.225∗∗
(0.027) (0.080)
Less than HS -0.010 0.134
(0.031) (0.080)
High School -0.026 0.182∗∗∗
(0.014) (0.032)
Some College 0.031 0.190∗∗∗
(0.020) (0.043)
Over 65 0.110∗∗∗ 0.201∗∗∗
(0.024) (0.053)
N 374.000 372.000
R-Squared 0.435 0.409
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
46 Table 16: Senate Election
Exhaustion Exhaustion Adj
Black 0.048 -0.378∗
(0.038) (0.175)
Hisp 0.054 -0.143
(0.057) (0.288)
American Indian/Alaska Native 0.013∗ 0.033
(0.005) (0.033)
Asian 0.018 0.147
(0.016) (0.091)
Less than HS 0.041∗ 0.191∗
(0.019) (0.097)
High School 0.020∗ 0.155∗∗∗
(0.009) (0.043)
Some College 0.030∗∗ 0.216∗∗∗
(0.010) (0.058)
Over 65 0.018 0.165∗
(0.013) (0.066)
N 374.000 373.000
R-Squared 0.128 0.156
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
47 Table 17: State Senate Election: Exhaustion and Vote Exhaustion Exhaustion Adj
Black -0.344 -1.106
(0.207) (0.570)
Hisp 0.533 -0.556
(0.317) (0.809)
American Indian/Alaska Native 0.076∗ 0.156
(0.036) (0.092)
Asia 0.029 0.296
(0.078) (0.180)
Less than HS -0.166 -0.021
(0.122) (0.334)
High School -0.033 -0.068
(0.074) (0.134)
Some College -0.195∗-0.278
(0.081) (0.163)
Over 65 0.016 0.076
(0.030) (0.101)
N 141.000 139.000
R-squared 0.139 0.150
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
48 Table 18: State House Election: Exhaustion and Vote
Exhaustion Exhaustion Adj
Black -0.443 -2.877∗∗
(0.325) (0.859)
Hisp 0.653 1.057
(0.448) (1.118)
American Indian/Alaska Native 0.258 1.117∗
(0.168) (0.518)
Asian -0.289∗∗ -0.474
(0.098) (0.344)
Less than HS -0.258∗-0.429
(0.127) (0.387)
High School 0.049 0.018
(0.032) (0.118)
Some College 0.154∗∗ 0.345∗∗
(0.047) (0.123)
Over 65 -0.073 0.451
(0.063) (0.228)
Candidates 0.035∗∗∗ -0.032
(0.009) (0.024)
N 91.000 91.000
R-Squared 0.316 0.288
Standard errors in parentheses
Regression models
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
49

Redondo Beach City CouncilJuly 30, 2024

Redondo Beach City Council
July 30, 2024