No spoilers with approval voting?

[Rob Wilson]

I read the following highlights of approval voting on the CES page today:

• More expressive
• No vote splitting or spoilers
• You can never get a worse result by voting for your favorite
• Significantly fewer spoiled ballots
• Ballots look the same, except the rules indicate that you may vote for any number of candidates
• Results are still easy to understand: a simple list of the candidates along with how many votes they received
• Tends to elect candidates who would beat all rivals head-to-head
• Tends to elect more moderate winners
• Alternate candidates get a more accurate measure of support

Saying that there is no vote splitting is arguably true, but no spoilers? No voting system is spoiler free.  Wouldn't it be better to say minimizes spoilers?

[Toby Pereira]

Also, on this page - https://electology.org/spoiler-effect - it says:

"In mathematical terms, the spoiler effect is when a voting method exhibits failure of a property known as independence of irrelevant alternatives.

Spoilers are possible in all ordinal (“ranked”) voting methods, but not in Score Voting (aka Range Voting). That includes the simplest form of Score Voting, called Approval Voting."

However, approval and score voting do not, in practice, pass independence of irrelevant alternatives, even if they do in some theoretical sense. At least some voters give their approvals and scores based on the candidates standing, not in isolation.

You also get the Burr dilemma - http://rangevoting.org/BurrSummary.html - which can result in a certain amount of vote splitting and can cause the spoiler effect.

So I would argue that it should say it reduces vote splitting and spoilers rather than eliminates them.

[William Waugh]

On Monday, August 8, 2016 at 2:47:44 AM UTC-4, Rob Wilson started the conversation that you can find at https://groups.google.com/forum/?fromgroups#!topic/electionscience/CyLDf8RcaTk .

Approval Voting is Range Voting. Conclusions made about one apply to the other. At https://groups.google.com/forum/?fromgroups=#!topic/electionscience/uIzIoBSK-wc I report on a mathematical experiment that convinces me that these systems can exhibit a spoiler effect if the voters do not use strategy to ward off the spoiler effect.

William Waugh

[Clay Shentrup]

Take a stack of Score Voting or Approval Voting ballots.

Remove a non-winning candidate from consideration and re-tabulate.

The winner will not change.

This is not true for any ranked method.

[Toby Pereira]

It's a nice property, but it doesn't have much real-world relevance on its own.

[Kevin Baas]

agree with the others - there technically are no spoilers in approval voting for a single-winner election.

if it's a multiple winner election, however, since it's a first-past-the-post system (no vote transfers), you get a lot of wasted votes, which leads to a very non-proportional result.  (highly skewed towards the majority)

[William Waugh]

What is the difference between a spoiler and vote splitting?

Context at https://groups.google.com/forum/?fromgroups#!topic/electionscience/CyLDf8RcaTk

[Rob Wilson]

Under this criteria, plurality would be considered as spoiler proof.

If every voter only judged each candidate individually without regard to any other candidate in the race, I'd agree that approval voting is spoiler proof. Realistically though, that isn't going to be how a significant portion of people behave.  If a voter doesn't like any of the candidates, he is at the very least going to "approve" of the guy least objectionable (assuming he bothers to show up).

[Rob Wilson]

Personally, I'd consider vote splitting as disenfranchisement of a voter who would vote for two or more candidates, but can only choose one. I'd say a spoiler is a non-winning candidate whose mere participation changes the outcome. Since approval voting gives you the option of voting for all the candidates you like, then maybe it can be considered immune to vote splitting. Since the addition of a candidate can influence whether or not you vote for a different candidate though, approval voting is still not immune to spoilers.

[Clay Shentrup]

On Wednesday, August 10, 2016 at 11:11:00 AM UTC-7, Rob Wilson wrote:

Under this criteria, plurality would be considered as spoiler proof.

I somewhat dispute that. You could think of Plurality as "rank all the candidates and the one with the most first-place votes wins". In which case, it still fails.

In that model, Plurality Voting in practice is just a case where the ballot is truncated. The same way "Ranked Choice Voting" is a 3-limit truncated IRV.

Or put another way, your argument says that IRV passes IoIA if we truncate the ballot to a single ranking.

[Jameson Quinn]

By the same token, you can think of approval as "vote a score ballot, then use the largest gap as the approval threshold", and then approval still fails this property.

I think that approval clearly reduces spoilers, but it does not make them impossible. When I want to claim something in this neighborhood, I say approval "means voters can always avoid spoiled elections", by which I mean, if a ballot set is a stable equilibrium, then removing a candidate not in the Smith set doesn't change the outcome.

--
You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.