The Effect of Ranked-Choice Voting in Maine
Ranked-choice voting (RCV), or broadly the ability to rank candidates according to voter preferences, has been a topic of close interest for several years. Maine was the first state to use RCV in federal elections, starting with the congressional election in 2018. This was expanded to include the first-ever ranked-choice presidential primary in 2020. After being used twice, we can begin to inquire into its impact on the voters and the political process in Maine.
What impact does RCV have on voters, especially compared to “traditional” plurality voting? Similarly, what impact has RCV had on campaigns in Maine? This post details some of the answers to these questions in upcoming research funded by the New Initiatives in Election Science grant, administered by the MIT Election Data and Science Lab, and provided by the Hewlett Foundation.
What is “Ranked-Choice Voting”?
Ranked-choice voting has served as a catch-all term for several different types of voting systems that use ranked ballots, even though each of these systems has its own unique set of properties and types of representation.
These can be broken down into two main types; instant-runoff voting (IRV) and single-transferrable vote (STV). IRV is currently used in Maine and turned down in a statewide referendum in Massachusetts in 2020. It works by allowing voters to rank as many of their preferences as they see fit. If a candidate breaks 50% of first-place votes, they win the election. If not, it goes into a vote reallocation. The candidate with the fewest votes is eliminated, and voters who named that candidate as their first are then allocated to their second choice. This process continues until one candidate reaches a minimum 50%+1 of the remaining votes.
In an STV system (such as the one used by Cambridge, Massachusetts, referred to as “Plan E”), there are multiple winners in each race, and voters rank as many candidates as they wish. If a candidate breaks the threshold of first-place votes to be elected, votes from different precincts are randomly assigned to their second choice. This continues until all available seats in an election have been filled. In effect, this is a form of proportional representation and carries distinct representational effects separate from the actual rankings of candidates.
How does RCV impact voters?
To answer this question, I took a two-pronged approach. The first study consisted of an experiment alongside a survey of Maine voters conducted following the 2018 election, the first to use RCV in a statewide election. The second study consisted of examining observational data on campaign civility, campaign spending, and voting behavior in the 2018 midterm election in Maine.
In the experiment, I examined the impact of both RCV and the number of candidates on voter satisfaction, voter confidence, perceptions of partisan benefit, ease of use, and the time it takes to cast a ballot. Using a survey experiment fielded through Lucid Academic, I randomly assigned respondents to vote using ranked-choice voting or plurality voting on a randomly chosen slate of 4, 8, or 12 candidates.
The findings of the experiment were stark. I found that RCV produced significantly lower levels of voter confidence, voter satisfaction, and ease of use. It also increased the perception that the voting process was slanted against the respondent’s party. Similarly, I found that it increased the amount of time it took to vote by nearly 12 seconds per candidate than voting using a plurality ballot.
On a positive note for RCV reformers, I found that voting for non-major party candidates — which reformers tout as “sincere voting” instead of voting for a candidate simply due to their realistic chance of winning — was increased by 5 points among the respondents who were given the RCV ballot instead of the traditional ballot.
In the second part of my study, I fielded a survey of registered Maine voters in the fall of 2019 to examine their opinions about ranked-choice voting. The results were slightly more encouraging than what was found in the experiment. I found that fewer voters found the RCV ballot to be difficult to fill out. However, this may be a sign of social desirability bias; I also found that a majority of respondents were concerned with how “other people” were more likely to be confused and spoil their ballots, which may be a way of signaling that the respondents themselves were confused by the ballot design. Similarly, the survey measured a significant drop in voter confidence compared to results in the 2016 Survey of the Performance of American Elections that cannot be attributed to the “winner’s effect” alone, which is the tendency for voter confidence to drop in voters who supported the losing side.
RCV’s impact on campaigns
In the second study, I examined a series of claims about the impact of RCV on campaigns and civility. One of the main claims made by reformers about RCV is that it will make campaigns more civil, as campaigns will have an incentive to seek the second-place vote of supporters of different candidates. To study this claim, I first conducted a difference-in-differences analysis on independent expenditures for and against candidates. In this analysis, I found that negative spending increased significantly in Maine following the implementation of ranked-choice voting, casting doubt on the claim that RCV makes campaigns more civil. To provide more evidence, I also created a dataset of all Facebook advertisements that mentioned any congressional candidates for 2018, the first year that RCV was used in Maine. I then conducted a sentiment analysis to find each advertisement's sentiment (whether it was negative or not). I then used genetic matching to approximate an experiment to find the impact of RCV on civility. In doing so, I found that the 2018 campaign was even more negative than in paired districts around the country.
Another claim made about RCV is that voters will be more likely to express their sincere preferences instead of only voting for less-preferred candidates who have a greater chance of winning. Using the congressional vote returns gathered by MEDSL for 2018 and 2016 and using a basic difference-in-differences framework, I found that non-major-party candidates saw a 6 point increase in vote share due to RCV, which is very similar to the 5 points that were seen in the experimental results.
What are the effects of RCV?
In my research on ranked-choice voting and its impacts on individual voters and the political system, I’ve found it seems to boost non-major-party vote share (and increase “sincere voting”), but it does not have most of the behavioral and campaign effects that reformers tout as potential benefits of the system. I should note this study is not equipped to analyze the long-term downstream effects of RCV on representation, policymaking, or its use in executive elections. These outcomes may be impossible to determine. However, as the research on election reform has demonstrated repeatedly, it is often impossible to predict the actual outcomes of election reforms (either by experts or those who advocate for or against them). As such, ranked-choice voting will need to be the subject of study in election science and political behavior for years to come.