9

Approval voting is a single-winner electoral system where each voter may select ("approve") any number of candidates. The winner is the most-approved candidate.

On paper, approval voting seems like an excellent system:

  • It is easy to explain
  • You can use the same voting machines and ballots
  • It never requires more than one round of voting
  • It avoids spoilers and reduces the opportunity for strategic voting
  • It encourages candidates to be more centric in their policies

So why isn't approval voting used more often in elections?

JonathanReez
  • 50,757
  • 35
  • 237
  • 435

7 Answers7

10

Approval voting does not meet the following criteria:

  • Majority: if a majority of voters prefer one candidate to every other candidate, does that candidate win?
  • Majority loser: if a majority of voters prefer every other candidate over one candidate, does that candidate lose?
  • Mutual majority: if a majority of voters prefer every member of a group of candidates to all candidates not in the group, does a member of that group win?
  • Condorcet: if there is a unique Condorcet winner, does that candidate win?
  • Condorcet loser: are Condorcet losers guaranteed to lose?
  • Smith and ISDA.
  • Later no harm. If I add a candidate to my ballot, is that guaranteed not to make a more favored ballot lose?

Later-no-harm and Condorcet are incompatible, so no voting method supports both. But many voting methods manage one or the other.

You can use the same voting machines and ballots

Ballots sure. I'm not convinced on voting machines. But I'm also not convinced that using the same ballots is a good idea. What if people don't realize that they are supposed to change how they vote and only approve one candidate. Then it devolves into plurality voting.

It never requires more than one round of voting

But that means that voters are choosing among a large number of similar alternatives. Two rounds of voting mean that people who don't know who to choose in the first round can still participate in the second.

It avoids spoilers and reduces the opportunity for strategic voting

It fails later-no-harm, so there are spoilers and strategic voting.

The suggested way to do approval voting is to either order all of the candidates and pick those you prefer more than you wish to avoid or to pick a pivot and compare the rest to that person. If you're ordering all the candidates, you might as well do ranked voting. And choosing a pivot is definitely strategy.

The biggest problem that I have with it is that it doesn't allow me to say that I prefer one candidate to another except in the approval. So I can basically say that I prefer one group of candidates to another group. But I only get two groups. I would rather use one of the more complicated ranked voting methods as better expressing my actual preferences.

Brythan
  • 89,627
  • 8
  • 218
  • 324
  • 8
    Some (all?) of these criteria only make sense in the context of ranked ballots, which Approval is not. Applying ranked criteria to non-ranked systems requires you to make assumptions about what's going on inside voters' heads, which is controversial. See Majority criterion:Approval voting, for example. – endolith Jun 01 '18 at 13:58
  • 4
    Also, these criteria presume that a majority getting their way at the expense of a minority is a good thing, while advocates of cardinal voting systems believe the opposite (see Peter Emerson's books on how majoritarianism causes war, for instance). Anyway, I doubt any of these academic reasons are why it's "not used more often", which has more to do with unfamiliarity among the population. – endolith Jun 01 '18 at 14:00
  • 1
    @endolith all voting uses some simplification based on assumptions of how people think or mater to arrive at a group opinion. So far as I know there is no science to say how much error there is in any theory of thinking, so all systems are backed only by argument; do you like the kinds of fairness X system offers? This answer shows some well regarded fairnesses this system does not offer. I'm also pretty sure these aren't actually closely related to why laws haven't changed, but I think it is very useful information to anyone interested in the idea. –  Jun 01 '18 at 14:53
  • 4
    @notstoreboughtdirt I'm specifically talking about these ranked-ballot mathematical criteria, though, and the assumptions you need to make in order to shoehorn a non-ranked system into them. (For instance, "The candidate who received the most votes was not the legitimate winner because, in their heads, a majority preferred someone else.") These assumptions are subjective, not objective, so it's not cut-and-dry whether these voting systems meet these criteria. (And these criteria aren't objectively good, anyway.) – endolith Jun 01 '18 at 16:10
  • 1
    That's pretty much what I mean by kinds of fairness. This is judging a fish by climbing, but if climbing is what you are interested in it's good that there are established ways to judge. –  Jun 01 '18 at 17:29
  • @endolith Note sure I am following your argument but it seems to me that the only assumption is that people (or at least some of them sometime) rank candidates (i.e. have more specific preferences than approve-disapprove). This is certainly an assumption about “what goes on in people's heads” but a pretty mild one and I am not sure it should be controversial. Importantly, people express these kinds of preferences all the time. – Relaxed Jun 02 '18 at 07:03
  • 5
    @Relaxed Or people have ratings in their heads, which are "flattened" into preferences when using ranked ballots, etc. "The majority judgement experiment proves that the model on which the theory of social choice and voting is based is simply not true: voters do not have preference lists of candidates in their minds. Moreover, forcing voters to establish preference lists only leads to inconsistencies, impossibilities and incompatibilities." https://hal.archives-ouvertes.fr/hal-00243076/document – endolith Jun 02 '18 at 20:10
  • @endolith The paper is intriguing, thanks for sharing it, but I do not find it particularly convincing. It would be a bit long to discuss in comments but as far as I am concerned, the authors go way beyond what their data actually show. The fact remains that people's judgments (if you object to the word “preferences“) are more complex than approve-disapprove, even in this experiment. This in turns means that a pure approval voting system is open to some of the paradoxes pointed out by Brythan. – Relaxed Jun 02 '18 at 20:39
  • 1
    There's a last point that's missing: in practice most voters continue to vote for only one candidate. (Source: I work for an electronic voting company, and most approval voting contests we run exhibit this in spite of the UI making it explicit that one can pick as many candidates they'd like.) – Denis de Bernardy Jun 08 '18 at 07:08
  • 1
    Later-no-harm is a terrible criterion which is incompatible with Favorite-Betrayal. I'd go so far as to say that satisfying Later-no-harm is itself a negative trait. – eclipz905 Aug 08 '18 at 20:37
  • 1
    @notstoreboughtdirt The linked voting system criteria are useful heuristics, but overly focusing on them misses the forest for the trees. What we should care about is finding a system that maximizes the satisfaction of the entire electorate. Doing that requires experimental data to confirm the theory. For this, I suggest looking into the voter satisfaction index score. – eclipz905 Aug 08 '18 at 20:39
  • 1
    @DenisdeBernardy You have that kind of data? Is it published anywhere? It should be – endolith Aug 30 '18 at 13:51
  • @endolith: the data is private, and I've moved to new projects since, but my understanding is it's fairly common knowledge in the voting industry. FairVote might have interesting data for you if you ask. Try their outreach department (Ruben or Scott @). At the very least they'll know where to send you to get data. – Denis de Bernardy Aug 31 '18 at 05:25
  • 2
    @DenisdeBernardy FairVote is very biased against approval voting and would love to use your data to attack it, if it really shows what you said. :) There's other data showing that 2 to 3 approvals per ballot is typical, though. https://sites.google.com/a/electology.org/www/pirate-elections-germany http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.734.5743&rep=rep1&type=pdf http://scorevoting.net/DartmouthBack.html Statistics about distribution of number of approvals per ballot from your employer would be very helpful in the debate. – endolith Aug 31 '18 at 14:05
  • @endolith: see this link, which references a primary source whose findings were consistent with the data we were seeing with my former client - https://www.fairvote.org/the_troubling_record_of_approval_voting_at_dartmouth – Denis de Bernardy Sep 01 '18 at 18:08
  • 1
    @DenisdeBernardy Yes, I'm aware of FairVote's 2 favorite examples. I'm looking for more data from the real world and unbiased sources. It sounds like you have some, and it would be helpful if it were published somewhere, even if it backs up FairVote's claims. – endolith Sep 01 '18 at 18:10
  • 3
    @Relaxed: "Psychological research has shown that cardinal ratings (on a numerical or Likert scale, for instance) are more valid and convey more information than ordinal rankings in measuring human opinion.[33][34][35][36]" from https://en.wikipedia.org/wiki/Cardinal_voting#Analysis There's even a quote there that Arrow apparently conceded later in life that his assumption (or restricting analysis to ordinal systems) was probably wrong. YMMV. – the gods from engineering Apr 17 '19 at 23:18
8

Don't discount the weight of history. The generic answer to a closely related question, "why is voting system X used in this election" is frequently that the voting system is picked by the winner, and the winner likes system X because he or she just won under it. Approval voting was only formalized relatively recently, so it hasn't had long to become a favoured system.

origimbo
  • 21,139
  • 2
  • 54
  • 78
5

Probably the biggest reason in the US is just unfamiliarity; people are used to thinking in terms of binary Us-vs-Them (elections, sports tournaments, etc.) so the concept that there could be other candidates that are not as divisive, but more acceptable overall, is foreign. We're used to the majoritarian concept of trying to beat the other guys into submission, rather than trying to find a consensus/compromise that we're all ok with.

The biggest psychological problem with Approval voting is that it's not as expressive as other systems. A candidate that you really love and a candidate that you merely find tolerable both get an equal, full-strength vote from you. Ranked and rated systems both allow more expression of preferences.

Probably because of this, in elections that have used Approval, most voters only voted for their favorite candidate, treating it like a FPTP election.1 2 Some places that previously used it repealed it for this reason.

endolith
  • 3,551
  • 20
  • 39
4

It is worth pointing out that there are people trying to make a strong case for using approval voting much more broadly, see, e.g., this book: https://www.springer.com/us/book/9780387498959 by Brams and Fishburn. They at least would probably argue that it's not for any fundamentally good reason that it is not being used, rather just due to a combination of historical accident and lack of insight. Of course, others will disagree.

present
  • 445
  • 1
  • 5
  • 7
3

It encourages candidates to be more centric in their policies

This is not a reason for, this is a reason against. Selling a voting system as skewing the vote plays well with very few people.

Another reason is because your actually voting for 'least disliked' rather than someone that anyone wants,so who is actually going to support the proposal? You're selling a negative.

Yet another reason is, unless there is very strong reasoning, people do not change existing stable systems.

-1

It used to be (sort of)

You can ask the U.S. Congress why they did away with single ballot, multiple candidate voting via the 12th Amendment to the U.S. Constitution in 1804.

Prior to that point, the Constitution required the following for presidential elections:

The Electors shall meet in their respective States, and vote by Ballot for two Persons ... they shall make a List of all the Persons voted for, and of the Number of Votes for each; ... The Person having the greatest Number of Votes shall be the President ... In every Case, after the Choice of the President, the Person having the greatest Number of Votes of the Electors shall be the Vice President. (Article II, Section 1, Clause 3)

While this isn't all-candidate approval voting (you can't vote for everyone you like), it does share the very valuable attribute with vanilla approval voting that a vote for one candidate is not necessarily a vote against another, mitigating the spoiler effect. It can be considered a hybrid of approval and plurality.

The objections based on metrics such as majority preference, etc. are a little bit cherry picked. As noted, it is impossible to satisfy every conceivable metric as "optimal" using a single voting strategy or any combination of strategies. Approval voting maximizes voter satisfaction more so than almost any other method

A fundamental problem of voting is that popularity is at odds with preference. Approval voting sides with popularity over intensity of preference, wherein a candidate or position that is favored even slightly by a large number of people is given more weight than a candidate or position that fewer people intensely want to win. Other methods purport to honor preference or ranking among candidates, but are more prone to the spoiler effect due to weaker respect for the tolerability of a choice (plurality offers no respect for approval beyond one's "top" choice, which means second-best and worst are treated equally). Range voting is a compromise between these extremes that retains considerable versatility.

As with every method we have seen before, it comes with a tradeoff--Do you want your most liked candidate to win, so badly that you are willing to pit perfectionistic tendencies against each other and throw the race to the mediocre coalition with the greatest solidarity, or do you want to arrive at consensus about which candidate the greatest number of people think is tolerable? You can't have both.

And of course some methods just outright forfeit information on both preference and popularity, and are easy to manipulate and game, which is the problem we are trying to solve.

Given that the original Constitution's voting system yields sizeable benefits we expect from Approval Voting, by reducing spoilers and eliminating partisan polarization on single tickets, I think investigating the rationales of the Congressmen who did away with it would be a worthwhile inquiry.

pygosceles
  • 791
  • 5
  • 11
  • Comments have been moved to chat; please do not continue the discussion here. Before posting a comment below this one, please review the purposes of comments. Comments that do not request clarification or suggest improvements usually belong as an answer, on [meta], or in [chat]. Comments continuing discussion may be removed. – CDJB May 30 '23 at 09:55
-3

Australia (and possible New Zealand... it was a Kiwi who introduced me to the concept) does have Instant Run-Off Voting which sounds like this. The basic idea is that if you have five candidates A, B, C, D, E. You can rate them in 3, 2 4, [n], 1 indicating the order of preference for candidates you want.

The results are tallied and the lowest scoring candidate is then eliminated and votes for said candidate are redistributed to the next available candidate on the list until one candidate has 51% of the vote. So if E was the lowest my vote would go B, then A, then D. the [n] is to represent that I absolutely cannot stand Candidate D and would never elect that way so he was never rated. I do not want D (sorry for the pun).

The problem is... well... get a good read on the Australian election system cause it is some bizarre voting regulations... essentially it is not unheard of for Aussie to have a general election ballad that has upwards of 100 candidates for one open seat that only one will win (you don't have to mark all in order, may will just go down the party preference list and order them that way).

The downside of this system is that until the introduction of computer counting machines, this system required several recounts that extended the knowledge gap between election night and the concession speech.

hszmv
  • 16,062
  • 2
  • 29
  • 53
  • 4
    Instant Runoff Voting /the Alternative Vote has concepts of ordered preference and of round voting which approval voting lacks, which modifies the selection of traits it might be nice for a voting system to satisfy which the two systems meet. – origimbo May 31 '18 at 19:40
  • Just FYI, from an Aussie who knows the electoral system from the inside... First, IRV is different to approval voting (which the Q is about); it's more complex but has (theoretically) better outcomes. The 100+ candidates ballots happen only for Senate elections, where multiple winners are selected (and thus the threshold is below 50%+1). Also, nowadays one can rank the whole parties, i.e. groups of 2-10 candidates, which reduces the list to "mere" 20-30 :) Single-winner ballots (for the lower house representatives) are usually 6-9 candidates long. – Zeus May 26 '23 at 04:15
  • ...Counting is still mostly manual, but there are techniques to obtain ~95% of the result overnight (for the more important lower house elections). – Zeus May 26 '23 at 04:18