Ten Critiques (And Defenses) On Approval Voting
Posted by on September 21, 2018 at 6:43 PM
By Aaron Hamlin, The Center for Election Science. Reposted from ElectionScience.org
Dig into voting methods and it won’t be long before you find criticism on different approaches. Approval voting (choose one or more, the candidate with the most votes wins) is no different. Unfortunately, analyzing voting methods—even when the methods themselves are simple—can get complicated.
To further complicate matters, we know through Arrow’s Impossibility Theorem that ranking (ordinal) methods must all fail some basic criteria that we’d prefer they not. Another theorem, the Gibbard-Satterthwaite theorem, shows that no voting method (beyond dictatorship or lottery) is immune to tactical voting.
Where does that leave us when evaluating a voting method?
In practice, we have to determine what the best fit is. We can set criteria to pass, run computer simulations, sort through available election and polling data, and look at practical considerations for implementation and use. It’s a combination of these approaches that leads us to push for approval voting in single-winner government elections.
We like approval voting for these elections because it:
- Is simple (in understandability, ballot design, implementation, and auditability)
- Favors consensus candidates
- Encourages competition through more candidates
- Gives otherwise invisible candidates get a more accurate and clear reflection of support
- Enables voters to always support their honest favorite
- Removes vote splitting almost entirely, which addresses the spoiler effect
- Avoids unpredictable anomalies such as non-monotonicity (this is when supporting a candidate can hurt that candidate and opposing a candidate can help that candidate)
- Minimizes voter utility loss when minor anomalies do occur
That said, approval voting has its detractors. Here are some anti-approval voting arguments (in their best form) alongside rebuttals.
Critique #1: Approval voting hurts voters who approve of additional candidates beyond their favorite (Hence, it violates the later-no-harm criterion).
Specifically, the later-no-harm criterion means that by supporting additional candidates, a voter can't hurt the performance of those more-preferred candidates. Strangely, a voting method that passes this criterion can still hurt a voter who simply adds candidates. This is commonly overlooked when considering whether a voting method passes later-no-harm. See our later-no-harm article for more on this nuance.
With this clarified, approval voting can in fact violate later-no-harm. Approving an additional, less-preferred candidate creates direct competition with a more-preferred candidate. Each candidate you approve is pushed forward by the exact same amount. This potential later-no-harm violation is tough to completely avoid because of how approval voting works. You can make expressions on multiple candidates and that information is calculated and considered simultaneously. Keep in mind that this criterion violation is only a potential and not a guaranteed outcome.
Inadvertently hurting a candidate on your ballot can sound frustrating, but the upside is that approval voting still always lets you support your honest favorite candidate(s). That support for your favorite will never hurt you. Conversely, many other voting methods that pass later-no-harm can trip up here, creating opportunities for voters to hurt themselves when supporting their favorite.
Also, any potential loss from later-no-harm that occurs with approval voting is not necessarily at the group level. In fact, these compromises at the individual level (even when they violate later-no-harm) can help the electorate as a whole to support the consensus-style winners favored by approval voting. For example, a moderate liberal who supports both a liberal and moderate candidate helps to support a more consensus winner in the same way as a moderate conservative who takes the same compromise strategy.
Many voters should likely take this compromise strategy. That’s because failing to do so could mean that they get a winner who's the polar opposite of their preference. Further, if supporting a slightly less preferred candidate does in fact lead to having that candidate surmount their favorite, the utility loss is not severe. That is, the candidate who wins is not significantly worse than the preferred candidate.
Compare that to other voting method failures that cause more catastrophic outcomes. There, outright bad candidates can win. And that catastrophic failure with other methods can occur at both the individual and the group level.
Consider also that later-no-harm is only a potential issue with approval voting when a voter's favorite and compromise candidate are close in an election. When this closeness is not the case, later-no-harm is not an issue because there’s no risk of hurting your favorite. Those who find issue with this suggest that risk-averse voters will only choose one candidate to avoid this situation. But supporting only one candidate doesn’t always make sense. And when it does, the tactic would really never makes sense for every voter, particularly in government elections.
Critique #2: Approval voting degenerates to bullet voting (choosing only one candidate).
Some who take approval voting's later-no-harm violation seriously suggest that the best tactic is for voters to choose only one candidate. That is, to bullet vote, the equivalent of regular plurality voting.
What we tend to see is that the average number of votes per ballot increases as the number of candidates increase. That is, fewer candidate options tends to mean that voters choose fewer candidates, and more candidate options tends to mean that voters choose more candidates. This is intuitive.
In fact, in many cases where there's an approval voting election, a large number of people do tend to bullet vote. The important part is that voters have the option to support multiple candidates when they need it—even when it's only a fraction of voters who need it. And it only takes a fraction of voters who support multiple candidates to change the outcome of an election. Even in an extreme situation where 90% of people bullet vote, those 10% choosing multiple candidates can sway the election (most often for the better).
Being able to support multiple candidates is a rather big deal for third parties and independents. As those candidates' support grows, it's essential that voters be able to support them to give their ideas credibility when it's warranted. It’s preposterous to suggest that the same third-party sympathizers who hold their nose while voting for a major party wouldn't support both a major party and a third party if given the chance. There is nothing to lose for them and everything to gain.
For a look at cases where it makes sense to bullet vote and when it doesn't, look at our approval voting tactics page. You'll find that while there are cases when it makes sense to bullet vote, there are essential cases where voters need the opportunity to approve multiple candidates to cast a more meaningful ballot.
Critique #3: Approval voting is inexpressive because you can only approve or disapprove of any particular candidate.
Firstly, approval voting is far more expressive than plurality voting, the status quo. Nothing can be less expressive than plurality voting. Nothing.
Here, we’ll define expressiveness as the number of meaningful ways (permutations) that you could cast a ballot. Let’s look at how four types of ballot expressions (including approval) stand up with three and eight candidates.
Here, ranking methods are assumed to allow only three rankings because that’s the way it’s been in practice due to ballot complexity. Both scoring methods (using four intervals of expression) and approval are also corrected by subtracting two permutations. We do this because supporting all the candidates maximally or no candidates at all is a mathematically meaningless ballot. This calculation approach gives a more conservative permutation count for approval voting. Still, there’s an argument that casting approvals for no candidates or all candidates allows the voter to say something meaningful even if it doesn’t change relative candidate standings.
Some are bothered that approval voting doesn’t let you differentiate candidates in more than two groups—approve/ not approve. It’s not the same expression as a numbered ordering or multiple scores. Approval voting has you either approve each candidate or not. But when you have more candidates, approval voting makes up for that flexibility by letting you say yes or no for all the candidates rather than just three.
A subtlety here is that, depending on the ranking voting method, not all the information is necessarily used. For instance, a method that simulates sequential runoffs (instant runoff voting), only looks at first-choice preferences in any particular round. The remainder of each ballot is ignored, and may be completely ignored throughout the entire tally depending on which candidates are eliminated. The number of candidates is also a factor. For instance, when more than eight candidates run, approval voting has more expressiveness than ranked methods, according to its number of ballot permutations.
Scoring/grading methods are the clear winner in expressiveness here. Using intervals of only four, this creates flexibility for simple ballot design yet still provides expressiveness.
All said, approval voting remains competitive in expressiveness among other classes of voting methods.
Critique #4: Approval voting is too novel.
Of all the single-winner voting methods, approval voting has the most similar feel to the status-quo (plurality voting). All you’re doing is changing the ballot directions from “vote for one” to “vote for one or more”. The votes are even tallied the same. That similarity is convenient considering approval voting’s potential impact.
It’s not possible to switch the voting method for single-winner elections with a smaller technical change than what approval voting can.
Critique #5: Approval voting violates “one person, one vote” and is unconstitutional.
Approval voting does not violate “one person, one vote”. The term “one person, one vote” refers to the weight of votes, not to how votes are expressed.
The U.S. Supreme Court made the “one person, one vote” rule explicit in Reynolds v. Sims (377 U.S. 533). The rule stated that no vote should count more than any other so that it has unequal weight. This unequal weight would violate the Equal Protection Clause of the Constitution. And it was Baker v. Carr (369 U.S. 186) that extended the Equal Protection Clause to districting issues. In Reynolds, the state of Alabama set up its districts so that they varied wildly in population. The districting was so bad that it gave some voters’ ballots as much as 41 times more weight than others. Because the weights of the ballots were different between districts, that violated the “one person, one vote” rule.
Even someone who votes for multiple candidates doesn’t get an unfair weight for their ballot. Say there’s a six-way election and both of us are voters: you vote for one person; I vote for the remaining five. Despite voting for five times as many candidates as you, all of our choices are tied. I didn’t get an advantage over you by choosing more candidates.
Consider also that voters are already allowed to vote for multiple candidates in at-large races using a method called bloc plurality. For instance, a city council may simultaneously elect five representatives. Some voters may vote for five candidates while others only vote for one or two. Here, we have an example of a currently existing method that allows expressions of multiple candidates and there is no constitutional issue present.
In all, because each voter’s ballot has the same weight, approval voting doesn’t violate “one person, one vote”.
Critique #6: Approval voting wouldn’t make any difference in the winner, or if it did, then the winner would be bland.
Changing the voting method within an election doesn’t necessarily change who wins. For instance, US President Ronald Reagan won in 1984 by more than 17 percent of the vote than his closest opponent. He would have won with virtually any voting method. Indeed, the same is true for any election where the winner takes more than half the votes (assuming only one round).
Approval voting has the potential to change the election outcome in any election where the winner had less than 50% of the vote. And really, approval voting could change the outcome in even more elections—by changing who would have run. Candidates who otherwise wouldn’t have run could now enter the race, unafraid of being spoilers. And one of those new candidates could win.
Approval voting could further influence the outcome by changing the dynamics of the election. Encouraging multiple candidates and less incentive to sling mud, candidates could be forced to talk more on the issues. This could change how voters perceive them and and persuade voters to cast their support differently.
Another outcome change that approval voting provides is the reflection of support for losing candidates, particularly third parties and independents. These candidates get a more accurate reflection of support under approval voting because voters don’t have to fear wasting. While this doesn’t necessarily change the winner, it still changes the election dynamics because less-known candidates can be given a fairer shake.
As to whether the winner would be bland, this doesn’t hold. Candidates who don’t take strong stances or don’t take stances at all are unlikely to be acceptable to many voters. A consensus candidate of the kind approval voting would prefer is not the same as a candidate who fills the lowest common denominator.
Critique #7: Approval voting doesn’t elect a majority winner.
No voting method can guarantee a majority winner when there are more than two candidates running. Indeed, the concept of majority is challenging and counterintuitive within voting methods. Runoffs, for example, merely contrive a majority by artificially eliminating candidates (who may be superior) so that only two candidates remain.
The concept of a majority in voting methods is complex enough to warrant its own article, which we’ve written.
Critique #8: Approval voting hasn’t been used in modern government elections.
Technically, it has, though the governments using it may not have known they were using approval voting. In the US, approval voting was used in an advisory initiative for Oregon. Other countries have used approval voting to make ballot question decisions. Even quasi-governmental organizations like the UN have used approval voting, though the public doesn’t vote in that election. Interestingly, Greece did use approval voting to elect its legislature from 1864-1926 before transitioning to proportional representation.
What is true is that (aside from Greece) voters haven’t yet used approval voting to elect a candidate to government office. But that doesn’t mean it shouldn’t be used for that purpose. Academics have been laying the theoretical groundwork for approval voting since the late 1970’s. There’s normally a time gap between when a voting method is formally studied and its implementation, so it is not so strange that it hasn’t yet been used to elect candidates to government seats.
While approval voting hasn’t yet been used by voters to elect candidates, we have a plan to change that.
Critique #9: Approval voting is unworkable for government and other competitive elections.
Plurality voting is the voting method that is unworkable for competitive elections. It’s awful in nearly every way from vote splitting, polarizing winners, the spoiler effect, and the barrier to entry it creates for alternative candidates and new ideas. The top of this article lists all the benefits of approval voting. These benefits unambiguously outclass the status-quo plurality method we use now.
Our government makes critical decisions from which laws operate within our lives to how enormous sums of money are spent. We currently decide who makes those decisions using a terrible voting method (many experts would agree that it’s actually the worst voting method). Approval voting is vastly better than plurality voting and should absolutely be used instead. We want approval voting instead because we want to have better and more representative people making these important decisions that shape the world around us.
Changing the voting method in government elections to approval voting is not only the right decision, but it has potential for enormous real-world impact.
Critique #10: Approval voting isn’t as good as voting method X, so voting method X should be used instead.
To say that there’s controversy over what the "best" voting method is (even just within single-winner methods) would be an understatement. Voting theorists debate the topic passionately. (Though a collective of international experts has given a preference for approval voting.) When evaluating a voting method we must often look at multiple metrics (sometimes involving math) and this can make for daunting discussion.
These metrics include several areas:
- How much utility on average do we expect a voting method to provide to its voters?
- How easy is the voting method to understand?
- How easily does the voting method lend itself to procedures such as hand counting and audits?
- How likely are unacceptable anomalies?
- How well does the voting method reflect support for all the candidates?
- How much competition is the voting method likely to encourage?
- How fair is the voting method at reflecting support for losing candidates?
Based on these factors and the current environment for voting methods, approval voting is arguably the best fit for single-winner government elections. It tends to pick good winners, gives all candidates their due, and is super simple.
It’s also important to consider that while some methods may technically perform better, other factors such as complexity play a role. There’s a curve of diminishing returns in utility that can occur as complexity increases. So while a particular complex method may offer slightly better results on average, there can be a high complexity cost to pay. With approval voting, however, the utility payoff is quite large with only a small (if any) cost for complexity.
All voting methods have their issues, but we have to settle on something for any particular election. Cities and states don’t even have to use the same voting method, so there’s opportunity to experiment. Given the appalling status-quo, approval voting actually presents less risk than doing nothing. The expected value from approval voting far surpasses this absurdly low bar.
Approval voting has its flaws, but those flaws are by no means fatal, particularly in comparison with other options. Approval voting has been used in government decisions before and has been used in large organizations as well. Between that usage and its academic track record, approval voting has been adequately vetted. And it is unquestionably better than what we currently use.
Here, we’re mainly talking about approval voting in the context of single-winner government elections. Admittedly, in a different environment where voters are savvy to alternative voting methods, more sophisticated methods may be worth the added complexity. But in our current environment, most people don’t even know what a voting method is. Also, it’s worth noting that a method being more complicated doesn’t necessarily mean that it performs better.
In the meantime, even when we take approval voting through the grinder, it still comes out as a solid choice. And the longer we wait to use it, the longer we have to put up with the ills created by our current method: choose-one, plurality voting.
Disclaimer: The views expressed in this blog post and these essays are strictly those of the author and do not represent the views of the Bridge Alliance Education Fund, the Bridge Alliance, or the Bridge Alliance’s member organizations. Additionally, the Bridge Alliance Education Fund makes no representations as to the accuracy of this post’s contents.