Voices & Choices

Rate of “Bullet Voting” Depends on Candidate Strength, Party Cues, and Other Factors

Rate of “Bullet Voting” Depends on Candidate Strength, Party Cues, and Other Factors

Many factors likely influence how a voter uses their rankings in a ranked choice voting (RCV) election. As of 2021, the median portion of voters who choose to use multiple rankings on their ballot is 68%. In this post we briefly discuss two factors that affect rank usage: voters’ perceptions of candidate strength and a candidate or party messaging encouraging or discouraging use of rankings.

 

A Candidate’s Front-Runner Status Impacts Whether Their Voters Rank Backup Choices

There is an intuitive feeling behind the idea that supporters of front-runner candidates might use fewer rankings, knowing rankings beyond their first are unlikely to be reached because their top choice is unlikely to be eliminated. In exploring this idea, we examined the relationship between front-runner status and rank usage in 196 past single-winner RCV elections in the U.S. All these elections had three or more candidates. Given our data limitations on tracking front-runner status prior to election day, we used winner status as a proxy for front-runner status. We did this knowing that in some cases winners come from behind in later rounds as well as the fact that new front runners can emerge throughout a campaign. In each election we grouped ballots by whether their highest-ranked candidate was the winning candidate or any other candidate. We then analyzed two metrics for each of these groups of ballots, mean valid rankings used and rate of voting for only one candidate, also known as “bullet voting.”

On their own, the distributions of each metric lend support to the idea that supporters of front-runners are less likely to use more rankings. In 78% of elections, the supporters of winners had lower mean rank usage than supporters of other candidates. Likewise, in 73% of elections, the supporters of winners had a higher bullet voting rate than supporters of other candidates.

We paint a fuller picture of the distribution of the two rank usage metrics above by improving our measurement of front-runner status by including the winners’ first round margins as a variable. We do this based on the hypothesis that large first-round margins correspond to candidates who are more strongly perceived as front-runners prior to the election

Analyzing the relationship between a winner’s first round margin and their supporters’ greater bullet voting rate shows that the two are correlated, with a 10-percentage-point increase in first round margin corresponding to a ~3.5 percentage point average increase in bullet voting rate differential.

A similar trend appears present when comparing a winner’s first round margin and their supporters’ mean rank usage.

Overall, this historical analysis provides some reason to expect that in elections with a highly preferred candidate, the candidate’s supporters are aware of their candidate’s lead and will likely rank less than supporters of candidates which have a greater chance of being eliminated.

 

Case Study: Re-Election Voting Patterns for 4 California Mayors

To better understand how voter behavior changes based on the context of an individual election, we will examine the four cases in which the same mayor was elected twice under RCV and for which we have enough ballot data to determine voter ranking behavior. The mayors are Ed Lee from San Francisco, Pauline Cutter from San Leandro, Libby Schaaf from Oakland, and Jesse Arreguin from Berkeley. 

In each case, the candidate was first elected when there were at least four candidates on the ballot in a race that took multiple rounds to determine a winner. When the candidates were each reelected, they won first-round majorities in elections that had fewer challengers, indicating their strong position as popular incumbents. 

Jurisdiction

Year

Winner

Candidates

Rounds

Avg ranks for winner

Avg Ranks Overall

Winner's 1st rd pct

San Francisco, CA

2011

Ed Lee

16

16

2.45

2.57

31%

San Francisco, CA

2015

Ed Lee

6

1

1.57

1.99

56%

San Leandro, CA

2014

Pauline Cutter

4

3

2.22

2.23

49%

San Leandro, CA

2018

Pauline Cutter

4

1

2.28

2.3

52%

Oakland, CA

2014

Libby Schaaf

16

15

2.68

2.59

29%

Oakland, CA

2018

Libby Schaaf

10

1

2.08

2.29

53%

Berkeley, CA

2016

Jesse Arreguin

8

4

2.37

2.32

49%

Berkeley, CA

2020

Jesse Arreguin

4

1

2.22

2.42

63%

When these mayors were first elected, voters who selected them as first choice ranked between 2.22 candidates (Cutter in 2014) and 2.68 candidates (Schaaf in 2014). During their reelection bids, voters for these candidates typically used fewer rankings on their ballots than the voters who first elected them four years prior. The most drastic decrease is for Ed Lee, whose 2011 voters ranked 2.45 candidates and whose 2015 voters ranked 1.57 candidates, a 36% decrease. In three of these four cases, voters for the incumbent front-runner decreased their ranking usage compared to four years prior. 

This change in ranking behavior supports the hypothesis that voters use more rankings when they believe the rankings will matter. In a crowded race with an uncertain outcome, voters tend to rank more backup choices. In an election with a popular incumbent likely to win a first-round majority, voters rank fewer candidates on average.



Candidate Attitudes Impact Ranking Behavior

Candidates who publicly oppose ranked choice voting also appear to impact the behavior of their votes, perhaps due to voters absorbing their top-choice candidate’s views about RCV. 

For example, former Maine Representative Bruce Poliquin (R) was a critic of RCV after Maine voters approved it at the ballot box in 2016. During Poliquin’s re-election bid in 2018, he narrowly lost to Democrat Jared Golden. Poliquin’s voters ranked an average of 1.8 candidates in a field of four, fully 25% fewer rankings than voters for Golden who ranked an average of 2.4 candidates. Voters for each of the two independent candidates averaged 2.7 rankings. 

Other RCV critics include Don Perata in the 2010 Oakland mayoral election and Mark Andrew in the 2013 Minneapolis mayoral election. In those cases, voters for the anti-RCV candidate used 6% and 8% fewer ranks than the overall averages. 

In the examples above, low ranking usage by voters for anti-RCV candidates did not impact the outcome because each of the three candidates earned a spot in the final round. However, it is instructive in understanding how candidate attitudes towards RCV may impact the behavior of their supporters. 

 

International Evidence Suggests Voters Follow Party Cues About Ranking

Recent research by Benjamin Reilly of University of Western Australia explores whether Australian voters tend to follow party cues when determining how to rank their ballots. Australian political parties often provide “how to vote” cards with a recommended ranking order, which sometimes include recommended rankings for candidates from other parties. Reilly notes that the proportion of “single rankings” by Labor party voters reached 72% in elections for which the Labor Party’s campaign material recommended single rankings. The rate fell sharply when the Labor Party’s materials instead recommended additional rankings. 

For more, see Ranked Choice Voting in Australia and America: Do Voters Follow Party Cues?

 

Conclusion: Voters Rank When it Matters… Unless They’re Encouraged Not To Rank

Based on the evidence above, we conclude that voters for less-favored candidates are more likely to rank back-up choices, while voters for established front-runners feel less of a need to do so. Additionally, candidate cues or party cues can supplement or detract from voters’ natural inclinations when choosing how many candidates to rank. 

 

Join Us Today to Help Create a More Perfect Union