Voting Methods Explained: How Ballot Lab Measures What Works
Author: Aaron Hamlin
Every election uses a voting method, whether voters notice it or not. The voting method determines what voters are allowed to say on the ballot, how that information is counted, and how many candidates are elected. Change the method, and the same voters can produce a different president, a different legislature, or a different governing coalition.
That is why Ballot Lab spends so much of its attention on voting methods. Voting methods are the machinery between what voters want and who ends up with power. When voting methods fail, the public’s preferences get distorted.
But comparing voting methods is harder than it first appears. Plurality, approval voting, instant-runoff voting, single transferable vote, party-list proportional representation, and other systems are not just different ways of counting ballots. They make different tradeoffs. They fail in different places. And they often aim at different goals.
The question we struggle with now is how we should evaluate voting methods and consider those tradeoffs. And to evaluate, we must measure.
Single-Winner
For a long time, academics have evaluated voting methods using a criterion-based approach. Someone comes up with a rule/criterion that they think a voting method shouldn’t violate. If you’ve heard of Arrow’s Impossibility Theorem, that’s the same idea. The theorem helped earn Kenneth Arrow the Nobel Prize. It proved that there was a set of reasonable criteria that voting methods would always violate in at least some cases. And while Arrow’s theorem only applied to ranking methods, every voting method misbehaves in some manner.
The problem with this criterion-based approach, however, is that it has limits and complicates meaningful comparisons. We can come up with endless criteria, whether it be whether a voting method punishes the voter for favoring their preferred candidate or, in some cases, punishing the voter for showing up at all. But this says nothing about how frequently a violation occurs, how much worse the outcome tends to be, or how important the particular criterion is in the first place.
We could instead look at how good a voting method was at electing a good winner. Often we think of a good winner as being whoever gets the majority. The challenge with that is that no voting method can guarantee a majority winner when there are more than two candidates. By using a runoff, whether explicitly or inferred through rankings, we merely contrive a majority. When we arbitrarily reduce the candidates to two, the remaining majority is just an illusion, merely a product of removing the other candidates.
There is, however, another kind of winner: a max utility winner—basically the candidate liked most by the electorate. We can approach this in multiple ways. In real, high-stakes elections, we can ask people to express their honest opinions about the candidates. We call this an honest-assessment measure, much like a control measure. Using this information, we can identify this max-utility winner. If we ask those same people how they’d have voted under different voting methods, we can see how close those voting methods were at capturing people’s true preferences.
Working with multiple members of the Ballot Lab advisory board, we explicitly carried out that experiment during the 2016 election. Capturing this kind of information to compare methods is critical. The downside is that it’s time and resource intensive to collect and analyze this data. And while the data is great, it’s still only one data point. And any election we don’t timely gather information for, we lose that opportunity forever. Ballot data alone isn’t rich or honest enough for us to really see how voters felt.
This approach to evaluating voting methods is important. We need to be scientific. If we say a voting method is measuring some construct, say candidate support, then we need to measure that as directly as we can. And when we measure that support, we should be able to compare it to what the voting method said. In this way, we can ask questions about voting method accuracy in a way that’s falsifiable. So a voting method that’s not accurate will fail to mirror the honest assessment measure.
A separate approach is to step back from real elections. We can use computer simulations to approximate countless election scenarios to see how well different voting methods do, on average, at capturing this high-utility winner. While we get far more data this way, we’re limited by the assumptions in our model. This approach can give us a broad initial look at how voting methods perform, but it’s helpful for us to double-check our assumptions with real-world data—the experiments drawn from actual elections.
Another factor here is that in either case (via experiments through survey sampling or computer simulations) we’re accounting for strategic voting. Strategic voting is when voters are dishonest about their vote to improve the election’s likely outcome. When people say how they’re voting in a survey, they’re allowed to be strategic. So a voting method’s performance, including when voters are strategic, is already baked in. And when we look at simulations, we can allow the model to include strategic voting.
Multi-Winner
Our evaluation process is more difficult when we use methods that elect more than one candidate simultaneously. That’s because we’re trying to do more than just maximize utility. Our max-utility winner candidate is one that finds their way to the ideological center and away from the extremes. If we were to max out only our utility metric in the multi-winner system, all our winners would be strong, but they’d also all be the same. At that point, we would ask ourselves why we want to fill so many seats if we’re just looking for one particular kind of winner.
What we’re also trying to do in multi-winner systems is to approximate the diversity of ideas from the electorate. So we’re getting representation not just from the middle but across the ideological spectrum. And we want to represent this in proportion to support that’s there. That concept is called proportional representation.
The challenge with this, which you may have guessed, is that at a certain point, you have to pick the metric you’re trying to optimize for or what kind of compromise you’d like to make. When we look at the literature, it suggests that we do indeed want to make a compromise and not just go all in on utility or proportionality.
So how do we measure proportionality? This is trickier than it appears. We want to measure how close an observed set of winners matches a best-fit outcome that reflects the electorate being maximally represented. So if 15% of the electorate supports a certain kind of candidate, they should get representation from those kinds of candidates in a similar number. And so on for all the other groups.
There is one well-regarded metric that gets at proportionality called the Gallagher index. The challenge is that this approach only works for a certain type of multi-winner method called a party-list system. In a party-list system, each voter picks a party, and the method allocates seats proportionately according to how many people support each party.
How do we compare proportionality across methods, not just party-list systems? Right now, we can’t. But it’s possible. It’s just a matter of taking a common measurement and applying a new index. This kind of measurement hasn’t been done before, but supporting our work at Ballot Lab helps ensure we make this a first.
One of the best predictors of proportionality is merely how many candidates we elect at once. It’s hard to represent the diversity of a population with just a few representatives. The literature recommends a minimum of five seats to achieve a reasonable level of proportionality. Again, though, we don’t have a good common metric. Also, even when we get the right metric, we need to learn the right balance between proportionality and utility.
Fitting It All Together
A big part of our work at Ballot Lab is to objectively evaluate voting methods. We can see how they perform and consider that alongside the complexity and logistics of getting them implemented. While doing all that is hard, we see that as our mission.
It’s imperative that we make sound decisions when we make voting method reforms. We look forward to your support in tackling these important challenges and ultimately helping to enable better voting method decisions.
Cite This Article
APA: Hamlin, A. (2026). Understanding and evaluating voting methods. Ballot Lab. https://ballotlab.org/ballot-lab-measures-voting-methods/
MLA: Hamlin, Aaron. “Understanding and Evaluating Voting Methods.” Ballot Lab, 19 May 2026, https://ballotlab.org/ballot-lab-measures-voting-methods/
Chicago: Hamlin, Aaron. “Understanding and Evaluating Voting Methods.” Ballot Lab, May 19, 2026. https://ballotlab.org/ballot-lab-measures-voting-methods/