Basically, vote for ALL of the candidates of whom you approve.
Whoever gets the most approval (votes) wins.
With a paper system such as we have in the UK, then if you don't like any of the candidates you can use a blank ballot paper to record your DISapproval, with the option that if the disapproval ballots exceed the winning total of approval votes then a new election should be called with a new slate.
Mathematically, for a single candidate election, it gets extremely close to the optimum Condorcet result.
Approval voting then is just a "restricted" subset of "relaxed ranked choice". That is, any preference expressed in an approval voting ballot can also be expressed in a relaxed ranked choice ballot. Similarly, relaxed ranked choice voting can be considered a restricted subset of range voting. And our traditional "single choice" voting can be considered a restricted subset of all of them.
And really you should separate the mechanism for expressing voter preference from the mechanism for picking a winner. Those are separate things that, in theory, can be mixed-and-matched to produce various voting systems.
So, clearly range voting is in some sense the best (or tied for the best) ballot, in the sense that it allows the voter to express a range of preferences that is a superset of the other systems.
But there is also the issue of ballot simplicity. Some people might prefer the simpler ballots of the approval system, or even the traditional single choice system. But since those ballots have corresponding "range voting" ballots, you can deploy a "range voting" election while allowing voters to use the kind of ballot they are comfortable with.
Indeed you can imagine the idea of a "progressive ballot" in which the voter starts of with a simple "single choice" ballot, and can optionally refine their preferences with an approval ballot, then a ranked choice ballot, then, ultimately a "range" ballot.
I've actually implemented such a thing .
 http://macd.tk/pollplace - Note, this is running on an underpowered test server not intended for public use, so be gentle.
Thank you for highlighting that. Too often the ballot and the aggregation method are conflated.
On https://modernballots.com/elections/zombies/vote/ I implemented 5-star ratings as the default. Everyone's familiar with this kind of rating and it allows indifference between two options.
I think Ranked versus Approval/Score is a proper comparison.
Ranked voting requires that you order candidates. That additional complexity is undesirable to proponents of Approval or Score Voting, such as myself.
You've suggested removing the requirement to order candidates and called it "Relaxed Ranked Voting." Removing the ordering requirement makes "Relaxed Ranked Voting" approximately equivalent to Score Voting (sometimes known as "Range Voting") . The difference is that the numbers are reversed. In Score Voting with a range of 0 to 100, the highest score—that is, 100—is given the the candidate(s) you like the most. In your "Relaxed Ranked Voting," I assume the best score would be a 1. This seems needlessly confusing to the voter. Just use a scoring system that everyone is familiar with from public high school, where 100 is the best.
Relaxed Ranked Voting seems additionally complicated by each elected office on a ballot having its own bounds (a three-candidate race would presumably permit scores of 1 to 3; a five candidate race would be 1 to 5; etc.) Meanwhile, Score Voting, once adopted, would use a consistent score range for all elected offices. For example, 100 is always best.
Am I missing something else about Relaxed Ranked Voting that makes you prefer it versus Score Voting?
I find it frustrating that IRV has more momentum than Approval and Score. I am worried that we will squander an opportunity to adopt a truly superior system such as Approval or Score voting  if we end up with IRV. I want to get rid of Plurality as much as anyone, but replacing it with IRV seems a half-success that will temporarily exhaust regular voters' tolerance of change.
As you point out, a reason so many of us on the Approval and Score Voting hype-train talk about Approval more than Score is that it's largely compatible with existing ballots. You just mark as many as you want. Super simple. In fact, I argue that Approval voting is simpler than plurality voting because it removes the "select only one" constraint.
Similarly, "relaxed rank" voting is also just a restricted version of score voting. You might argue that a paper version of the "relaxed rank" voting ballot wouldn't be as clear or intuitive to the average voter as an approval or score ballot would be. And I wouldn't necessarily dispute that.
What I'm saying is that instead of forcing every voter to learn and embrace a new ballot, you can let them vote as they always have and, optionally, allow them to "refine" or "supplement" their ballot by additionally filling out an approval ballot along with their traditional one. (If the approval ballot contradicts the original one it is discarded/ignored.) If the voter has chosen to fill out an approval ballot, they can subsequently further refine their ballot by filling out a score ballot.
I'm just pointing this out because it may be an option that is more acceptable to more people. In fact, you could add these optional "refining" ballots to existing elections with the understanding that they will have no effect on the election, but allow people to see how the election would have turned out had the "refining" ballots been counted. (Like, maybe John Kasich would have been president or something.)
If you go to the link I posted, you can login anonymously and try out these "progressive ballots" for yourself. If you register (no personal info required or anything), you can create your own polls if you want to experiment further. (Note that this is not a real site or anything. Don't rely on it, and don't use it for anything serious. Not yet anyway.)
The point is, whatever type of ballots you're using - traditional single choice, approval, ranked, relaxed ranked, or score/range - the ballots should all be converted to score/range ballots and evaluated by an optimal score/range evaluation method (probably Schulze or something like it).
This also means that the ballots don't even have to be all the same type. Each voter could choose the type of ballot they feel most comfortable with, knowing that whichever type of ballot they choose, their vote will count as much as any other voter.
However, I have some concerns. The way I see the two voting systems is like this:
Advantages approval voting:
- much simpler
- eliminates spoiler effect
- may give third-party candidates better chances in most elections
- reduces negative campaigning, since the winner would have to be "approved" by like 70% of the country in typical elections.
- the winner may be someone who was #2 on 80% of the country's wishlist. So 51% of country won't love the candidate, as they may with RCV or two-round voting systems, but also the other 49% won't hate the candidate (perhaps only 20% will). So from that point of view, it would be "better". But it would be someone most are just content with.
- I believe even the organizations supporting it admit that it would lead to "centrist" leaders. Perhaps in most situations a centrist is preferable, but what if the country has gone in the wrong direction for 2 decades, and it needs a completely new direction? Would a centrist still be enough? Or would the approval voting system and people pick exactly the guy that is willing to go in a different direction this one time? I'm not sure what would happen in this scenario.
- also reduces negative campaigning, because a candidate would need some of the opponent's voters, too, to rank them as #2 on their list.
- easier transition to multi-winner RCV system for state legislature and Congress - and this alone is much bigger than just using approval voting system. Proportional representation beats all single-winner voting systems, including approval voting
- it may eliminate spoiler effect, but other than that, it won't do much else to help third-parties. The main two parties would likely still be elected for a long time, at least until population's thinking about at least one of the the two parties changes in a major way
- a bit more complicated to understand how votes are counted by the average person, and a higher number of "lost" votes (I believe 5% or so, compared to about 1% or less for AV).
I believe both could also be used strategically - as in rank #1 the person you think is more likely to win with RCV, or only vote for the person that's more likely to win with AV, instead of multiple people.
So I would qualify the two as: AV would be a great improvement over FPTP, while RCV would be a moderate, but still well worth it and welcome improve, for the fact that it would eliminate spoiler effect alone. However, if single-winner RCV makes it much more likely that multi-winner RCV (STV) is also adopted for state legislature and Congress, then I would definitely choose RCV over AV, because the ultimate goal should be to adopt proportional representation in the US.
I think proportional representation coupled with a limit of $200 of individual donations and a ban on any other political donations would greatly improve democracy in the US, and these are the main changes Americans should fight for, if they want all of the other issues (as Lessig often says) to be solved as well. First fix democracy and change the system to a better one, so that the people that actually represent you get to pass laws in your favor for whatever issue you (the People) want.
That's an advantage, not a disadvantage.
IRV/RCV has one more critical flaw which makes the "lesser of two evils" problem worse: it completely ignores any later preferences on your ballot. If you list your preferred third-party candidate first, IRV/RCV ignores your preference for one first-party candidate over another. As long as your third-party candidate can't win, then your preference gets respected. However, when your third-party candidate reaches the tipping point, it's entirely plausible for your preferred first-party candidate to get eliminated first, followed by your preferred third-party candidate, letting your least preferred first-party candidate win.
That's quite plausible, because it's common for most voters for a particular third party to have the same preferred first-party, while first-party voters more commonly vote for only that party.
That creates an incentive for third-party voters to continue voting a first-party candidate at the top, because in IRV/RCV, only your top choice counts.
That said, hopefully it'll improve the ability to show third-party support, if not actually get third-party candidates elected.
Score Voting and Approval Voting are simpler and better.
It's true that voting for other candidates you approve of could take the election away from your favorite, but that's not what people usually mean with spoiler effect. As long as you do approve of the winning candidate, your vote isn't spoiled. In Approval Voting, there's no context where any form of voting, strategic or honest, will have you worrying that your vote helps someone you don't approve of.
Proportional representation is not an improvement. The effect is that you end up voting for a party instead of a candidate and you end up with all the hard party line behavior seen in Europe.
But it's also not true that you can't do multi-winner voting with approval voting. It's completely trivial -- you have fifteen candidates and five seats and the five candidates with the highest approval get seats.
The problem is, that has the same issue as using RCV with multi-winner elections -- it disenfranchises people. To have multiple winners you have to combine districts. Then all the candidates get chosen by what 51% of the voters in the combined district want when before 20% of the seats could have gone to a different party because those voters had majorities in their old smaller districts. Now they get nothing and have no voice at all.
> or only vote for the person that's more likely to win with AV, instead of multiple people.
That doesn't actually help you with AV. Not approving of the candidate you most want to win buys you nothing.
The failure mode of AV, such as it is, is that it requires you to choose between approving of your second-most favorite candidate or not, when doing that could both cause that candidate to win against your favored candidate and cause them to win against your least favored candidate, and you don't know which one without knowing how everyone else voted. But that's because Arrow's Impossibility Theorem is a bear.
You're describing party-list proportional representation, but the parent post was about PR-STV, under which voters rank candidates and don't vote for parties.
> But it's also not true that you can't do multi-winner voting with approval voting. It's completely trivial -- you have fifteen candidates and five seats and the five candidates with the highest approval get seats.
> The problem is, that has the same issue as using RCV with multi-winner elections -- it disenfranchises people. To have multiple winners you have to combine districts. Then all the candidates get chosen by what 51% of the voters in the combined district want when before 20% of the seats could have gone to a different party because those voters had majorities in their old smaller districts. Now they get nothing and have no voice at all.
Proportional representation (with or without party lists) would prevent minority rule and provide minority representation. I'm not sure your suggestion of multi-winner approval had either of these benefits.
Party lists make the problem much worse because it's de jure but any multi-winner system devolves into de facto voting for parties. More choices means that party branding plays a bigger role in voter choice, and even if a voter has a preferred candidate, they have to support that candidate's other party members in order for them to be effective once elected. Which makes it easier for parties to enforce party loyalty.
> Proportional representation (with or without party lists) would prevent minority rule and provide minority representation.
I don't think I'm adequately explaining what I mean by disenfranchise.
Suppose a factory is polluting the lake, or the old County Bridge needs to be repaired, or the the local police are corrupt.
If a party needs to resolve those issues to win the district then that's what they'll do. But if you put those voters in a big pot with a million other people who vote for parties based on things like abortion and taxes, nobody in the legislature is going to care anything about the local problems. So those people get disenfranchised -- nobody cares about them because no individual legislator is going to keep or lose their seat based on how people in only that one district vote.
> I'm not sure your suggestion of multi-winner approval had either of these benefits.
The point is that all multi-winner systems are bad. Legislators need to be elected by their own district or they won't actually care about those people.
If that's true, it isn't consistent with another supposed disadvantage of PR-STV: that under it competition between candidates of the same party undermines party unity. Either way, if you approve of democracy, why not majority rule and increased voter choice?
> Suppose a factory is polluting the lake, or the old County Bridge needs to be repaired, or the the local police are corrupt.
> If a party needs to resolve those issues to win the district then that's what they'll do. But if you put those voters in a big pot with a million other people who vote for parties based on things like abortion and taxes, nobody in the legislature is going to care anything about the local problems. So those people get disenfranchised -- nobody cares about them because no individual legislator is going to keep or lose their seat based on how people in only that one district vote.
What you want can be achieved with proportional representation. Effective local government (itself elected proportionally) with appropriate powers should address local issues.
> Legislators need to be elected by their own district or they won't actually care about those people.
Under FPTP, too many legislators are elected without a majority of votes cast in their electoral districts, thwarting democracy even at that level. The problem typically accumulates over a whole chamber; if you need a third of votes in an electoral district to win it, you only need a sixth of votes nationally to control the whole chamber.
Unfortunately it's possible to do both at the same time. The fact that the party can kick people out for not sticking to the party line doesn't mean there won't be friction when two members of the same party are campaigning against each other for a seat.
> Either way, if you approve of democracy, why not majority rule and increased voter choice?
"Majority rule" is the problem representative democracy exists to solve. With direct democracy (or anything that sufficiently approximates it) you get bad choices because if you put something on the ballot that will give +1 to 51% of people and -10 to 49% of people then it passes.
> What you want can be achieved with proportional representation. Effective local government (itself elected proportionally) with appropriate powers should address local issues.
The only way that could work is if each level of government was only in charge of things that affect each of its constituents uniformly, which is not possible. For example The Feds have to regulate pollution because pollution generated in West Virginia can affect New York, but that also means that the interests of people in West Virginia are different than the interests of people in New York and so they each need their own representatives.
And it goes all the way down the stack. Zoning regulations affect the whole city but affect different neighborhoods differently, so each neighborhood should have its own representative on the city council.
> Under FPTP, too many legislators are elected without a majority of votes cast in their electoral districts, thwarting democracy even at that level. The problem typically accumulates over a whole chamber; if you need a third of votes in an electoral district to win it, you only need a sixth of votes nationally to control the whole chamber.
I don't think anybody is arguing that FPTP is not terrible. Approval voting or range voting fixes it without having to combine districts.
Parties' candidates already work with each other under STV, so party disunity isn't an unacceptable problem.
Expelling politicians should only affect candidates' chances of getting elected if voters want it to. There is no vote-splitting under STV. STV doesn't discriminate against independents. Expelled politicians might take their votes with them if voters choose to follow them.
I don't mean to suggest that the problem of party disunity isn't real, only that it's outweighed by STV's advantages. However, I don't consider excessive party control under STV at all significant; the voters can decide whether to back rebellious candidates.
> "Majority rule" is the problem representative democracy exists to solve. With direct democracy (or anything that sufficiently approximates it) you get bad choices because if you put something on the ballot that will give +1 to 51% of people and -10 to 49% of people then it passes.
I don't consider government democratic if it isn't proportional, but I agree about the advantage of representative democracy over direct democracy. We need representative government to deliberate on our behalf and we need proportional representation to make it democratic. The voting system can do little more than choose between majority and minority; it can't protect against tyranny of the majority, but it can replace it with tyranny of the minority.
We need many civil society protections against the tyranny of the majority. One is subsidiarity.
>> What you want can be achieved with proportional representation. Effective local government (itself elected proportionally) with appropriate powers should address local issues.
> The only way that could work is if each level of government was only in charge of things that affect each of its constituents uniformly, which is not possible.
There's a difference between optimal and perfect. Unfortunately the best we can do is still imperfect.
> And it goes all the way down the stack. Zoning regulations affect the whole city but affect different neighborhoods differently, so each neighborhood should have its own representative on the city council.
Part of the problem is geographically dispersed interest groups. It's not only political parties; under any single winner system, ethnic groups, professions, LGBT people, the disabled, the poor, etc. are vulnerable to under-representation (or a clean-sweep, if they are numerous enough and distributed in a particular way).
> I don't think anybody is arguing that FPTP is not terrible. Approval voting or range voting fixes it without having to combine districts.
I agree that single-winner cardinal voting systems such as approval voting and range voting would be an enormous improvement over FPTP. I think Reweighted Range Voting (the proportional variant of range voting) would be an enormous improvement even over single-winner range voting. The reason I'm not calling for RRV is that I don't know a single example of its use to elect a government. Without that I don't see how RRV can be implemented in countries such as the US, Canada, the United Kingdom and Australia.
The largest two parties will do deals with all the minor parties to try to gather their second preferences on their "How to vote for us" cards and mail-out advertising.
And then independent candidates don't just have to be #1, they have to beat the major parties PLUS all the other independents who traded their second preference advice to the main parties...
That ends up being why the "two-party preferred" polls (which of the two main parties are you going to rank highest) end up being far more important than the primary vote (who will you pick as your #1 preference) in Australian opinion polling.
So it is a nice idea, and I do like that Australia has it, but for single-winner elections (rather than the Aussie senate where six candidates get elected from each state-wide vote), it perhaps makes it harder not easier for independents to win.
The preferences of minor party voters tend to scatter pretty randomly.
Woo hoo, represent Cambridge, Mass.
... our City Council and the Parks and Rec department of Minneapolis are the only two governmental bodies in the US that use Single Transferable Vote to elect a multi-body chamber.
* Each ballot is given an initial "weight" of 1.
* The votes on the ballots are summed for each candidate, thus obtaining that candidate's total score.
* The candidate with the highest total score (who has not already won) is declared a winner. (Note that the first winner is the same as the winner of an ordinary single-winner Approval Voting election using the same ballots.)
* When a voter "gets her way" in the sense that a candidate she approved of wins, her ballot weight should be reduced so that she has less influence on later choices of winners. To accomplish that, each ballot is given a new weight = 1/(1+SUM), where SUM is the sum of the scores that that ballot gives to the winners-so-far.
* Repeat until the desired number of winners has been chosen.
(paraphrased from http://rangevoting.org/RRV.html )
No, Approval Voting only makes any sense when approval has concrete meaning, as in a case wherected you are voting on a group activity, and voting "approve" on an alternative is also a binding commitment to participate if that option is chosen (or, conversely, voting not to approve is a binding waiver of participation.)
Approval voting is probably the only system that's worse than FPTP. Under approval voting, the candidate who is the least-objectionable wins, regardless of whether voters actually prefer them to other candidates.
So, you could very easily end up with a single-issue candidate being "approved" by the vast majority and winning, even though absolutely nobody would choose them over any of the other candidates.
> Mathematically, for a single candidate election, it gets extremely close to the optimum Condorcet result.
It's misleading to refer to a Condorcet winner as the "optimum" result - the Condorcet criterion is one criterion that an election method can satisfy, but it does not guarantee the "optimum" result by any other criterion.
It's also doubly-misleading to refer to the Condorcet winner right after advocating approval voting, which does the exact opposite.
This is true, but not necessarily a bad thing at all. Democracy is a proxy for violence to prevent violence. Someone that everyone thinks is merely okay sounds better to me than one that is liked by a third and hated and despised by a third.
All voting systems have pathological cases. The question is how serious they are and how likely they are. There's a tendency for people to look at one of these cases and think that it is disqualifying, but that's not the right approach. One should look at the whole picture.
Approval voting is a special case of range voting where there are only two values. For small numbers of voters, adding more values to choose from will help prevent the pathologies. For large numbers of voters, that's not necessary -- the statistics make it extremely likely that you get the right result.
Every AV voter has to decide where their approval threshold is. For a candidate that they're iffy on, are they going to vote for or against them? If everybody has exactly the same threshold, then yes, pathologies become more likely. But in reality the thresholds will have some distribution, such that a candidate that everybody is iffy on will tend to get voted for by about half the people. If no other candidate tops that, I think you'd be hard pressed to argue that AV returned the wrong result.
You're asking us to imagine a scenario in which there is a single issue on which one candidate's position is very popular, but despite that, no other candidate who takes that position on that issue is acceptable to as many voters. In the kind of multi-candidate race that AV is intended to encourage, that seems unlikely to me, at least if the race is one of any importance. And don't you think the other candidates are going to point out that the single-issue candidate is exactly that?
Single-issue voting is certainly a problem -- one we have already -- but I don't see that AV makes it worse; in fact I think it would help, a lot!, by allowing races in which both centrists and more extreme candidates run.
BTW I agree with you that the Condorcet criterion is overemphasized.
Approval Voting performs exceptionally well as measured by Bayesian Regret.
There's also a theorem that it elects beats all winners given plausible models of voter strategy.
Cofounder, The Center for Election Science
If voters disagree about who the "other candidates" are, I don't see that as a problem.
For RCV, the main objection is probably that it's more complicated. Then again, Americans use stuff like electoral college, and superdelegates, and caucuses, and coin flips, to decide elections, and in comparison to all of that, the RCV system seems easy (just rank candidates in order of preference). So RCV probably has the "easy" objection between the two.