Here's my problem with Approval Voting: Let's say you are left-wing, and there is a left-wing, centrist, and right-wing candidate running. Do you approve the centrist candidate? If you approve the centrist candidate, you may block the right-wing candidate, but you also may block the left-wing candidate. There is no clear and obvious way for the voter to translate their preference into an approval vote.

This isn't a problem. Numerous contentious Approval Voting elections and exit polls have been held with Approval Voting, showing voters can certainly cast a vote without being paralyzed by indecisiveness.

If YOU can't make up your mind, then just don't vote. Seriously. You can get Approval Voting and just not even show up and vote. You'll still get massively better election outcomes because of all the people who do show up to vote are using an objectively better decision-making algorithm. Yes, objectively better.

http://ScoreVoting.net/BayRegsFig.html

I didn't intend to suggest someone wouldn't vote because they are "paralyzed by indecisiveness." My concern is that people will vote sub-optimally due to poor information about other voters preferences and poor understanding of the strategy involved.

That's a valid concern. http://www.electology.org/score-voting-threshold-strategy argues that in approval voting an honest ballot would give you (by some measure) at least 91% of the impact of a perfect tactical ballot.

An honest vote in first-past-the-post, say for your awesome friend Bob for President, would give you approximately 0% return compared to the proper strategy of voting for the lesser evil of Clinton.

This is a common logical fallacy. I'll explain.

Suppose Approval Voting gives you a utility ("happiness") of 10 if you're tactical, or 9 if you're honest.

Now suppose Voting System X, in which a sincere and tactical vote are obviously identical (ignore the fact that this requires the system to use randomness), gives you a utility of 7.

Now do you _really_ want to take the latter system because there's no disparity in your happiness based on whether you knew the best strategy? Or would you rather take the system that gives you a better result?

More details here. www.electology.org/topic/tactical-voting

I'm sorry, but you've completely missed the point. Nobody is talking about the "disparity in your happiness based on whether you knew the best strategy."

Obviously, if the honest approval ballot give me 9 utility points, and the strategy approval ballot gives me 10 utility points that's better then the system that always give me 7 utility points.

However, it seems to me that the situation is much more along the lines that an honest approval vote gives me 2 utility points, and a strategic one gives me a utility of 9.

See, I hold to political positions that almost always put me in the minority. If I only approved those candidates that that I actually liked, I'd only approve candidates with effectively zero chance of winning. This would let my least favorite candidate win, thus granting me very little utility.

Hence, I'll probably take System X, even if approval voting might, in theory, do better if everyone applied the strategy correctly.

> This is a common logical fallacy. I'll explain.

No, it's not. I agree with you. (But perhaps didn't express that right.)

Actually, that discussion is about an honest score voting ballot, not an honest approval ballot.

Hmm, the author confusingly mixes topics.

Approval Voting is Score Voting on a 0-1 scale. I tried to be clear about the distinctions as they apply. If you can cite any specific sentences that you think could be made clearer, please let me know.

I think I confused myself there.

If you want your site to be usable for drive-by-quoting in an Internet discussion, perhaps you want to make two copies of that text, one specifically for Approval Voting and one for score voting in general?

PS Overall nice website, I am reading more.

Thanks. We've changed platforms a couple of times and have lost some pages and introduced some broken links and "CSS" issues along the way. We need more manpower to make it pretty, but no one cares enough about democracy to give us the millions to invest in such things. :)

This was the original site from years ago: https://sites.google.com/a/electology.org/www/approval-votin...

Ironically, my startup just got through YC and my fingers are crossed that we're successful enough for me to one day fund such reforms, so we'll never be threatened by a Trump again.

Congratulations!

I wonder what the right strategy would be? I guess start small and introduce sensible and simple voting everywhere you can:

- Shareholder votes in a company - Where to go for dinner - Vote for mayor in a small town - ...

Most people don't even know that there are different voting systems, yet alone that some of them are vastly better than others. (It's like watching people play Monopoly, when they are much better games around these days.)

For what it's worth, voting system changes typically result from a two-party system splitting: http://repositori.upf.edu/handle/10230/506. When countries end up with more then two parties, those parties suddenly have an incentive to institute better voting systems. That would suggest you'd have to take out the Republican/Democrat duopoly to change the US system, but perhaps Trump will do some good after all.

Interesting. They had some changes in party systems in the US over the course of the country's life, but did they change the voting system?

Britain also had some changes (with labour replacing liberal at the beginning of the 20th century).

I hail from Germany, which traditionally has a more proportional system. (It's a weird mix that seems to do reasonably well in practice. They could replace the first-past-the-post of the `Erststimme' with a score voting system, but the Erststimme isn't that important anyway.) They never had less than three parties in the federal parliament after WWII, even with a 5% threshold needed to move in.

The UK had three somewhat relevant parties in the 20th century (though only two relevant parties per county). Why haven't they changed electoral system? (And when they had a referendum about it, why did they suggest the dreadful IRV?..)

The USA has had changes of party systems, but never in a way that resulted in a lasting third party.

Certainly there are places like the UK, or my homeland, Canada, that have had three relevant parties for a while without changing the system. Although both UK and Canada have forces pushing for reform. The recently elected government of Canada gave an election promise to change the system, but, well, I'm not holding my breath.

But I'm not saying at all that any country with multiple parties will enact voting reform, but statistically, it seems that voting reform is often brought about that way.

I suspect IRV was brought to vote in the UK, because that was deal that could be reached between the Conservatives who didn't want change, and the Liberal Democrats who did. However, I will note that IRV isn't so bad when selecting members of a legislature. Approval/Condorcet elect compromise candidates, but we don't want to simply have a legislature of compromise candidates, but rather representative of how people voted. But some from of proportional system is probably better.

That's a common logical fallacy. What's important is not "how much utility did I lose by not casting the most optimal vote in Voting System X?" What's important is, "how much utility did I lose by not _using_ Voting System X?"

Explained here with a simple graphic to visualize it. www.electology.org/topic/tactical-voting

That's an interesting point. Yes, approval voting, like all other voting systems, requires voters to not only know their preferences, but also to play a little `game' (in the game theory sense of the word).

The optimal strategy on who to vote for depends not only on preferences, but also on what you know about the other voters. Polls help here.

The nice thing about approval voting is that, even though their might not be one single, clear and obvious way to translate your preference into your vote; but there is a clear and obvious family of ways to do so. And they have nice properties.

In-first-past-the-post, people often elect to give their vote to the lesser evil, instead of their preferred (third-party) candidate.

In approval voting, you will never have to give a compromise candidate more `vote' than any other candidate you prefer.

Ie all families of sensible voting strategies for approval voting look like this:

- sort candidates in order of preference - figure out a cutoff - approve anyone above the cutoff

The optimal cutoff depends on the peculiarities of the election (polls etc).

In practice, people manage this quite well intuitively. (I have run approval votings for small things, like which restaurant to take the team to.)

I like the idea but don't agree that all you do is vote above the cutoff. If there are 10 choices and I approve of 3, but I know my 7 and 9 are the only possible winners, I'll likely vote 1 2 3 7.

Your best strategy is to approve all candidates you prefer to the expected value of the winner. http://www.electology.org/score-voting-threshold-strategy

This works out well, tending to elect Condorcet winners. http://ScoreVoting.net/AppCW.html

Thanks. So the best strategy is even simpler!

So you do vote above the cutoff. But the optimal cutoff is really simple to find from the polls.

I don't think that works. Polls tell you the levels of support of the candidates, they don't give you the probability of each candidate winning which is what you need to figure out the expected value of the winner.

I guess it's back to the prediction markets (or fivethirtyeight) then.

I don't think other system require "playing the game" in the same way as approval voting. You can game any voting system by voting strategically, but approval voting seems unique in actually expecting people to have to vote strategically.

You can vote strategically in any deterministic voting method. There's nothing special about Approval Voting. People can vote honestly based on whether they approve (have a favorable opinion of) a candidate. Or they can be tactical and base the decision on probability of winning.

And in approval voting at least your strategy is very close to honesty.

Your strategic consideration only consists in choosing a cutoff.

> I don't think other system require "playing the game" in the same way as approval voting.

You must be joking? Almost no one ever votes their true preference in first-past-the-post. People usually vote for the lesser evil of the top two candidates, because any other strategy (like voting for who you'd actually like to win the election) just wastes your vote.

In any case, check out eg https://en.m.wikipedia.org/wiki/Approval_voting for other people's thoughts on the system. There's much more online about all aspects and pros and cons that we don't need to rehash.

Sorry, I tend to think of FPTP as being so messed up it doesn't qualify as a voting system. I was comparing approval voting to alternatives like Ranked Pairs/IRV/Borda etc.

Oh, OK. I am arguing for Approval Voting because it's the simplest voting system up from FPTP, and easy to explain.

Borda Count really invites the gaming, though. https://en.wikipedia.org/wiki/Borda_count#Potential_for_tact...

Borda is better than Plurality and IRV regardless of the amount of tactical voters.

http://ScoreVoting.net/BayRegsFig.html

Interesting. I'd have to look at that simulation.

What about cloning candidates? The simulation only looks at strategic voting.

The simulation uses random utility generators, which sometimes create situations where two candidates are highly similar and thus "partially" clones. The point is to model reality.

Compliance with "independence of clones" is in this property chart. http://scorevoting.net/CompChart.html

I get where you are coming from on Approval Voting. The simplicity of the system is very attractive.

Other systems might be slightly better. But Approval Voting is dead simple to explain, and you don't even need new ballot papers compared to plurality.

The only system that's generally better is Score Voting aka Range Voting. But you want to _avoid_ ranked voting methods.

http://ScoreVoting.net/NESD.html

I don't think the argument presented there works.

Consider the example with A, B, and C at the end. It claims that approval voting would probably elect C, but why? If the voters apply perfect strategy, B supporters won't approve C, and B will win. The same is true of range voting, where B supporters will give C a rating of zero.

It does not seem escapable that a majority voter block can dictate their wishes on the whole population by exaggerating their stated preferences.

As long as you need a majority, that's still very democratic? (If it was a plurality only, that would be harder.)

Yes, but the page in question is arguing that approval/rank voting is better because it doesn't always elect the majority's first choice.

I'm pretty convinced that for single winner elections approval voting is pretty great.

I haven't convinced myself when multiple people can get elected.

Yes, Approval Voting is simpler and radically superior to IRV. www.electology.org/approval-voting-vs-irv

Bayesian Regret calculations show IRV is actually THE WORST of the five commonly discussed alternative voting systems.

The best is Score Voting, but it's not as simple as Approval Voting. http://ScoreVoting.net/BayRegsFig.html

Clay Shentrup Co-founder of The Center for Election Science

Cool. Here's a better introduction to score voting [1]

The site you linked looks like it was built in 1998

[1] http://electology.org/score-voting

I wrote most of that page. :)