The effect of Ranked-Choice Voting on third party donations

Tufts Public Opinion Lab
5 min readMay 16, 2024


By Abby Sommers (‘25), Jack Maniaci (‘24), Lucy Morisse-Corsetti (‘27), Maria Cazzato (‘26)

Note: This is one of a series of guest posts from students in the Tufts Political Science Research Methods course.

Ranked-choice voting (RCV) is an election method that involves listing multiple candidates on a single ballot according to the voter’s preference. If a voter’s first choice doesn’t receive a majority of first-choice votes, this candidate is eliminated and the voter’s second-choice is counted in the next round, and so on. Proponents note that RCV is particularly useful for increasing the likelihood of a third-party victory. Therefore, RCV could be a solution to the persistent issue of polarization and gridlock in American politics. With the ability to list multiple candidates, RCV eliminates so-called “strategic voting”; voters no longer have to choose between a third-party candidate, whom they most align with, and a major party candidate, who has a better chance of winning ( Their vote is not wasted by voting third-party. However, a major factor left out of this equation is whether changing election laws to RCV would actually increase voters’ confidence in a third-party candidate enough to change their behavior. One way to measure this effect is to look at campaign donations: if third-party or independent candidates receive a larger portion of donations under RCV, it may signal a shift in voters’ confidence and an endorsement of its proposed solution to polarization. At the same time, we can examine other potential effects of RCV via this confidence proxy, including those on racial and gender representation. To observe these effects, we look at elections before and after New York City’s implementation of RCV for municipal primaries in 2021.

How we conducted our study

The data were pulled from the NYC Campaign Finance Board online portal which provided donation and spending data from past city-wide elections. Our analysis focused on private funds rather than public funds from the city, because it is a better proxy for voter confidence. Unfortunately, the NYC CFB did not provide party affiliation or demographic information for the candidates, so we coded it ourselves. Because NYC allows candidates to affiliate with multiple parties, we used three separate variables to identify whether a candidate affiliated as aDemocrat, Republican, and/or with a third party. To make the private funding data more comparable across years and elections, we recoded the absolute dollar amounts to percentages of the total donations in that given year and race. Further, we identified Andrew Yang as an outlier, and removed him from the analysis; he was the only Asian independent candidate in the 2021 mayoral election, and perhaps due to his national profile he received far more donations than any other independent candidate in the race.

What we found

We ran three separate regression models to capture the full scope of our data. The first was a regression of only independent candidates based on the proportion of total donations to each race, controlled for race, gender, election type (RCV or single-winner), the office, and any secondary affiliation with major parties. When looking at the outputs for this, the statistically significant coefficients were the offices the candidates were running for, as well as black candidates as compared to white candidates. All three of these had p-values below 0.05, leading us to be at least 95% confident in their results. In regard to race, being a black candidate led to 0.265% less fundraising as compared to white candidates. When considering the elected office coefficients, both showed a decrease in funding when running for lower office, with the borough president candidates having 0.464% less funding, and the city council having 0.5016% less. These patterns reflect that people pay more attention to mayoral races in general, and are therefore more aware of all of the candidates, even the “smaller” ones.

The second regression we ran was the same as the first with the addition of the interaction terms of election type with race and election type with gender. This means that there were some new statistically significant coefficients in this regression model as compared to the last one. Black candidates still receive a smaller proportion of donations as compared to white ones, with a statistically significant p-score of 0.02 (indicating a confidence interval of 98%) and a 0.345% smaller donation proportion. The office variable also still produced a statistical correlation, with borough president having 0.4678% less donations with a p-value of 0.032 and city council having 0.4877% less donations with a p-value of 0.0002. However, the interesting new additions come from the inclusion of interaction terms with the election type variable, as well as the interaction of the election type variable with women candidates on their proportion of donations. RCV predicts 0.2908% fewer donations as compared to single-winner elections, with a statistically significant p-value of 0.029. This, interestingly, rejects our null hypothesis in the opposite direction than we anticipated — independent and third-party candidates who were white men actually raised less funding RCV. The other new significant coefficient is the interaction of gender and RCV. The coefficient here in conjunction with the coefficient for RCV indicates that women raised about the same amount under RCV as they had under the previous system, only men did worse under RCV.

The third regression model we ran included the same variables and interaction terms as the second, but instead of looking at just third-party candidates, looked at all NYC election candidates from 2013–2023. When looking at this model, there are some new correlations present, as well as an increase in the correlation for the trends presented in the models that considered just 3rd party/independent candidates. The implementation of RCV predicts a 1.0602% smaller proportion of donations each candidate received, with a statistically significant p-value of 0.00037. This helps provide statistical support for our alternative hypothesis, but in the opposite direction we expected (in that RCV depresses third-party donations rather than uplifting them). When looking at other variables, black candidates receive a 1.12% smaller proportion of donations than white candidates (with a p-value of 0.000137), and now when looking at the entire pool of candidates, we can see that Hispanic candidates also receive 0.676% less donations than white candidates (with a p-value of 0.0494). Independents and third-party candidates overall received 0.9839% of a smaller proportion of funding than mainstream party candidates (p-value of 0.0045). When looking at the interaction terms, the only statistically significant one was the interaction of Black candidates with RCV, which appeared to lead to a 1.1339% increase in funding proportion with a p-score of 0.004465, making this increase statistically significant. Thus, it appears that RCV helped Black candidates raise more money than they had under the previous system.

Overall, our results indicate that the shift to RCV did not increase the share of donations directed to third party candidates. However, there is some evidence in support of the notion that Black candidates fared better under RCV. Of course, given that the shift to RCV happened between 2019 and 2021, it is not clear to what extend the Covid-19 pandemic and the Black Lives Matter movement may have had on our results.



Tufts Public Opinion Lab

The Tufts Public Opinion Lab (TPOL) is dedicated to studying contemporary controversies in American public opinion using quantitative data analysis.