Jill Stein promotes IRV in Rolling Stone

[Clay Shentrup]

We are in a damned-if-you-do and damned-if-you-don't situation right now, which should be fixed by a simple legislative reform — that could be passed right now, by the way, for anybody who is concerned about wanting to change this rigged political system: The state legislatures can simply pass ranked-choice voting, which gets rid of the fear factor. It is used in many cities around the country from San Francisco to Portland, Maine, and many in between, and in many countries around the world. It allows you to rank your choices instead of just picking one; you don't have to make your vote a gamble.

Most people are [voting out of fear] right now. The last CNN poll that showed the majority of Clinton supporters are not really supporting her, they're primarily voting against Donald Trump, and the majority of Donald Trump supporters are mainly voting against Hillary Clinton. What's wrong with this picture? We live in a system that tells us to vote against what we are afraid of rather than for what we believe in.

Read more: http://www.rollingstone.com/politics/news/green-partys-jill-stein-on-why-bernie-sanders-should-go-third-party-20160531

[Rob Wilson]

I would agree that IRV would get rid of the fear factor, but that would only be because people don't understand how voting for your favorite can hurt you.

[William Waugh]

Thanks for flagging this instance of erroneous advocacy regarding voting systems. I added a comment.

On Wednesday, June 1, 2016 at 3:55:02 AM UTC-4, Clay Shentrup wrote:

...

Read more: http://www.rollingstone.com/politics/news/green-partys-jill-stein-on-why-bernie-sanders-should-go-third-party-20160531

[Andy Jennings]

On Wed, Jun 1, 2016 at 6:04 PM, Rob Wilson <blahf...@gmail.com> wrote:

I would agree that IRV would get rid of the fear factor, but that would only be because people don't understand how voting for your favorite can hurt you.

Especially for the Green party!

Maybe I'm oversimplifying their politics, but to me it seems obvious to me that:

1) They'll take votes mainly from Democrats.

2) If they get strong enough then the Democrat will get eliminated.

3) Lots of Democrats will have put the Republican second, so the Republican will probably win.

Thus, the Green Party, with IRV, will cause the Republicans to win.  (Argument will ensue over whether the Green was a spoiler.  IRV advocates like to argue that anyone strong enough to make it to the final round can't be called a "spoiler".)

[William Waugh]

Unfortunately, the comment area doesn't show up anymore on the page.

[Kevin Baas]

Nope, you got it totally wrong.  That's not how IRV works.

Let's say there are 3 parties. D,R,G.  Now a party needs a majority to win.

1) So consider the possibility that "they'll take votes mainly from democrats":  Now either republicans already have a majority, in which case it doesnt' matter at all anyways, or they don't, in which case, having the fewest votes, the green party votes will go to the second choice - which, if we assume are "mainly from democrats", will all go to the democratic party, and the democratic party will win.  So either way, unless Greens get a majority -- in which case Greens win --, the party that would have won if the Green party didn't run a candidate will win.

2) "If they get strong enough then the Democrat will get eliminated."  if democrat was the smallest first choice, and nobody got a majority, then yes, democrat will get eliminated, AND ALL THOSE VOTES WILL BE TRANSFERED TO THEIR SECOND CHOICE, which, if your premise 1) is true, will almost all go to the Green party, and the Green party will win.

3) If instead the second choice of dems was more republican then green, such that the republican would win, well then that's the correct outcome: that's the outcome that most people preferred.  had the green party not run a candidate, republicans would have won.

so in any case the correct candidate is elected.

[Andy Jennings]

On Sun, Jul 17, 2016 at 12:39 PM, Kevin Baas <happy...@gmail.com> wrote:

3) If instead the second choice of dems was more republican then green, such that the republican would win, well then that's the correct outcome: that's the outcome that most people preferred.  had the green party not run a candidate, republicans would have won.

I don't know if there is a "correct outcome", but in scenario 3, if some of the Green voters had put the Dem first (or even forgotten to vote), then the Green would've been eliminated first and the Dem would've won.

This does make the Green a spoiler by some definitions.  It may make Greens and Dems shame the Green voters (before or after the election) for putting the Green first.  And they may shame the Green candidate for running.  It makes people afraid to vote their favorite first.

It's a pathology we would like to avoid as much as possible.

[Toby Pereira]

Obviously this sort of thing can happen under IRV, but I don't see any particular reason why Democrat voters are more likely to put Republicans second than the Green voters are. This is especially given the exaggeration strategy that people are likely to engage in (those that put one of Democrat/Republicans top are likely to put the other bottom).

I'm not generally convinced that IRV causes people to not vote for their favourite for this reason. If a small party gets big enough so that it can defeat one of the frontrunners, it's no longer really a small party. And why would a former small party be less likely to win in the final run-off than an already-big party?

By the way, I'm not disputing that IRV keeps two-party domination, but I'm not convinced that this is a reason. I think the exaggeration strategy comes from the ranked ballot, and you'd probably find it under Condorcet as well.

Condorcet voting is less likely to get undesirable-looking results once the election is analysed, but I'm not sure how much it would improve voting behaviour relative to IRV.

Approval and score on the other hand would hopefully do a better job.

[Kevin Baas]

if some of the Green voters had put the Dem first (or even forgotten to vote), then the Green would've been eliminated first and the Dem would've won.

This is good.  By the Green voters putting the Dem first they are expressing that they prefer this option over the alternative.

If they put the Greens first and then the Dems, they are expressing that they prefer that option over the alternative.

And, indeed, this difference in expressed preferences translates to a potential difference in outcome, in the same direction of the change.

This is called "pareto efficiency", and it's an essential characteristic of a fair election, in which each voter is empower with a voice in the outcome.

Whether other voters "shame" them for voting their preferences -- well that is a separate issue altogether.  That is a campaigning issue, not a vote tallying issue.

[Kevin Baas]

problem with approval and score, even for a single winner election, is they lack the ability to transfer votes, so it wastes a lot of votes - not everyone gets in equal voice.  this is a first amendment issue.

simply normalizing the ballots to 1 or using allocation voting (max n votes) would enable vote transfering, and thus fix this first amendment issue.  and i dare say the outcome would be fairer, too.

[Kevin Baas]

however, it occurs to me that if greens had more fist place votes then dems, but less than repubs, with IRV this could be problematic for dems and greens alike:

dems would be eliminated, their votes transfered to second place

which might give repubs a majority.

...despite either greens or dems being the condoret winner and/or borda count winner.

that is, the green party can still spoil ballots, if they have enough first place votes to drop the dem candidate but not enough to get elected.

which would give green voters ample reason to betray their favorite and vote dem instead of green.

[Clay Shentrup]

On Sunday, July 17, 2016 at 1:19:18 PM UTC-7, Kevin Baas wrote:

This is good.  By the Green voters putting the Dem first they are expressing that they prefer this option over the alternative.

They would have already been saying that with an honest ballot! E.g. Green>Dem>GOP already says that you prefer Dem to GOP.

This dishonest ballot actually expresses that they prefer the Democrat to the Green, which is a lie!

And this causes the appearance of not being electable to become a self-fulfilling prophecy, ergo two-party domination, and cash will have a greater influence on the political process.

This is not what you want in a voting system.

[Clay Shentrup]

On Sunday, July 17, 2016 at 1:23:10 PM UTC-7, Kevin Baas wrote:

problem with approval and score, even for a single winner election, is they lack the ability to transfer votes, so it wastes a lot of votes - not everyone gets in equal voice.  this is a first amendment issue.

You have this completely backwards. IRV ignores much of the data marked by certain voters but not others. E.g.

http://scorevoting.net/IrvIgnoreExample.html

This happens because IRV transfers votes. It considers only your top ranking, and ignores all other data you've marked down.

Score/Approval by contrast are perfectly fair. A voter who casts a vote that is the reverse of yours has a precisely equal (but opposite) effect.

[Kevin Baas]

You have this completely backwards.  Because IRV transfers unused (surplus or deficit) votes, it counts every ballot as exactly one vote.

Conversely, since range voting has no method of transfering surplus or deficit votes, votes in excess of that needed or which are wasted on a candidate that cannot win are essentially just that: wasted.

"A voter who casts a vote that is the reverse of yours has a precisely equal (but opposite) effect."

Firstly, this would not imply that all ballots are equal - only that precisely opposite ballots are equal.

Secondly, this is false.

Lets say there are to candidates.  A has 100 points, B has 105 points.  Now a ballot that's 10,0 would change the winner to A.  But a ballot that's 0,10 would not effect the outcome.  Now together they cancel each other out, yes.  But that does not imply that taken independantly they have the same effect.

In any case the closer the a score is to quote, the more likely any give point for can affect the outcome; the more "voting power" each point is worth.  Conversely, the further from that needed to win, the less the points effect the outcome.

So a person who votes for a compettive candidate, their vote counts a lot more than someone who voted for either a landslide winner or hell's snowball.

The votes are not equal - the ones that are for candidates more on the cusp count more, and the ones who vote for overwhelmingly popular or unpopular candidates have their first amendment rights abridged.

[Kevin Baas]

you should have continued reading.  my next comment i think it was i considered the problem of favorite betrayal.  which i think is what you're refering to here.

range/score voting suffers from the same problem.  worse, actually, since the rational choice is always bullet voting.

[Toby Pereira]

Score voting doesn't suffer from favourite betrayal, if that's what you're saying.

[Toby Pereira]

Nor is the rational choice always bullet voting.

[Kevin Baas]

Score voting suffers from favorite betrayal AND later-no-harm, and the rational choice is ALWAYS bullet voting.

[Nevin Brackett-Rozinsky]

Kevin, do you remember that time when we were discussing the voting system you yourself created, and I strongly recommended that you eliminate your habit of arguing against mathematical facts?

Consider yourself reminded for a second time.

The utility-maximizing strategy, hence the rational choice, under score voting is to fully support all candidates with a positive Myerson-Weber rating, and fully oppose the rest*. Myerson-Weber is an exaggeration strategy, so every candidate gets either the maximum or minimum score, and multiple candidates will often receive the maximum score so it is definitely not bullet voting.

Furthermore, score voting is—emphatically—completely immune to favorite betrayal. No matter what the situation, it is *always* correct to give maximum support to your actual favorite candidate in score voting.

These are not controversial points. There is no “debate” about them. They are proven mathematical facts.

Nevin

*Technically the rational strategy says a candidate whose Myerson-Weber rating is identically zero can be given any score. Fully opposing them is consistent with that and makes the description shorter.

[Clay Shentrup]

On Sunday, July 17, 2016 at 3:12:02 PM UTC-7, Kevin Baas wrote:

You have this completely backwards.  Because IRV transfers unused (surplus or deficit) votes, it counts every ballot as exactly one vote.

Wrong. E.g. in the Burlington election, IRV eliminated the Democrat before even registering that the Republicans favored the Democrat to the Progressive. Whereas IRV acknowledged the fact that Progressive voters preferred the Progressive to the Democrat.

Conversely, since range voting has no method of transfering surplus or deficit votes, votes in excess of that needed or which are wasted on a candidate that cannot win are essentially just that: wasted.

The concept of wasted votes doesn't make sense in single-winner elections.

"A voter who casts a vote that is the reverse of yours has a precisely equal (but opposite) effect."

Firstly, this would not imply that all ballots are equal - only that precisely opposite ballots are equal.

That is the only meaningful kind of equality you can ensure.

So a person who votes for a compettive candidate, their vote counts a lot more than someone who voted for either a landslide winner or hell's snowball.

This is ideally what you want, since some people prefer X to Y by a small amount, whereas others prefer X to Y by a large amount. 

[Clay Shentrup]

On Sunday, July 17, 2016 at 3:49:38 PM UTC-7, Kevin Baas wrote:

the rational choice is ALWAYS bullet voting.

Okay, you're just being a troll at this point. 

[Kevin Baas]

On Sunday, July 17, 2016 at 10:24:29 PM UTC-5, Clay Shentrup wrote:

On Sunday, July 17, 2016 at 3:12:02 PM UTC-7, Kevin Baas wrote:

You have this completely backwards.  Because IRV transfers unused (surplus or deficit) votes, it counts every ballot as exactly one vote.

Wrong. E.g. in the Burlington election, IRV eliminated the Democrat before even registering that the Republicans favored the Democrat to the Progressive. Whereas IRV acknowledged the fact that Progressive voters preferred the Progressive to the Democrat.

I'm not going to bother to look into this in detail.  There are some common errors in counting which I pointed out on an earlier thread.  two which deal with how eliminations are handled.  This may be a result of those errors.  in any case, votes are transferred, which prevent (when done correctly) excess & deficit ballots from going uncounted.

 

Conversely, since range voting has no method of transfering surplus or deficit votes, votes in excess of that needed or which are wasted on a candidate that cannot win are essentially just that: wasted.

The concept of wasted votes doesn't make sense in single-winner elections.

Granted.

 

"A voter who casts a vote that is the reverse of yours has a precisely equal (but opposite) effect."

Firstly, this would not imply that all ballots are equal - only that precisely opposite ballots are equal.

That is the only meaningful kind of equality you can ensure.

Not true,  And you are fully aware that this is not true as I expressed a different kind of equality in the previous comment: one person, one vote.

 

So a person who votes for a compettive candidate, their vote counts a lot more than someone who voted for either a landslide winner or hell's snowball.

This is ideally what you want, since some people prefer X to Y by a small amount, whereas others prefer X to Y by a large amount. 

No, you want everyone's vote to count as much the same as possible, and the result to be proportional.

As you mention earlier it doesn't much matter in a single winner election, but in a multi-winner election, it's everything. 

[Warren D Smith]

In IRV, it is commonplace that the algorithm TOTALLY IGNORES many of
the preferences expressed by many of the voters.

For example, if A is the winner, and your vote is A B C D E,
then the algorithm will not even look at, will totally ignore the "B C D E."

If however, I voted "C E D B A" then the algorithm would not
ignore anything I said. So some voters are largely ignored and other voters
are paid attention to, which seems pretty unequal.

For another example if the IRV final round is A vs B,
it might be that some other candidate C was preferred versus both by
large majorities, e.g. 80% prefer C>A and 80% prefer C>B.
The opinions of A-top voters about C-vs-B are ignored, simply not even looked
at by the IRV algorithms. IRV asked the A-top voters to write stuff
on that part of their ballot, which unfortunately confuses many of
them into thinking
their opinions matter, but they do not. But the C-top voters get to
express their
opinions about A-vs-B and *those* count. Quite unequal.
And this selective ignoring is exactly how it can be that C can lose
and IRV election to somebody the voters greatly prefer.

Also, it can be mathematically proven that
in a C-candidate election in the limit when C is large,
the IRV algorithm always ignores
asymptotically 100% of the information the voters write on the ballots.
For some voters it may read their entire ballot, but the total amount
the algorithm reads is a fraction tending to zero of the total amount
written by all the voters.


--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Kevin Baas]

This is false.

Properly counted (removing the 3 errors ive mentioned), each voter has exactly one choice paid attention to.

Which is teh same number of choices paid attention to as any other vote.

That's why it's called "single transferable vote", because each person gets exactly one vote, and that vote has the ability to get transferred if their vote is not needed or is for a candidate that can't win.

This ability to transfer their vote prevents it from getting discarded and counting for nothing.

[Kevin Baas]

regardless of whether or not i'm "just being a troll at this point", bullet voting is always the rational choice for range voting.

[Kevin Baas]

just re-read a definition.

by "bullet voting" here i mean rating everything at the extreme - e.g. 0 or 10.

since range voting fails later-no-harm, if the voter knows all the other votes, their best vote is ranking one candidate 10 and the rest zero, reducing the system to simple plurality.  however, lacking perfect information, they'd probably be better off doing something like half 10's and half 0's.

[Jameson Quinn]

2016-07-18 13:49 GMT-04:00 Kevin Baas <happy...@gmail.com>:

just re-read a definition.

by "bullet voting" here i mean rating everything at the extreme - e.g. 0 or 10.

since range voting fails later-no-harm, if the voter knows all the other votes, their best vote is ranking one candidate 10 and the rest zero,

Wow. In the span of two sentences, you just managed to completely redefine the term "bullet voting" twice.

When you find yourself doing something like this, it's probably time to stop arguing against others, and start arguing against yourself. Imagine that one week from now, aliens threatened to destroy the world unless you could write something that they could send back through a time machine to change your mind today. In other words, make a good-faith attempt to look at the strongest possible arguments for the other side, rather than just reflex-replying "nuh-uh".

 

reducing the system to simple plurality.  however, lacking perfect information, they'd probably be better off doing something like half 10's and half 0's.

On Monday, July 18, 2016 at 12:40:50 PM UTC-5, Kevin Baas wrote:

regardless of whether or not i'm "just being a troll at this point", bullet voting is always the rational choice for range voting.

On Sunday, July 17, 2016 at 10:25:05 PM UTC-5, Clay Shentrup wrote:

On Sunday, July 17, 2016 at 3:49:38 PM UTC-7, Kevin Baas wrote:

the rational choice is ALWAYS bullet voting.

Okay, you're just being a troll at this point. 

--
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 electionscien...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Warren D Smith]

KB has it wrong ("each voter has exactly 1 choice paid attention to").

IRV makes many choices -- one choice per round.
If the decision in some round is between X and Y
(which to eliminate) then voters who ranked X or Y top at that moment,
will be paid attention to at that moment. All other voters will have
no effect on the X vs Y choice -- even if 99% of the voters
ranked X>Y, they might all be ignored (if, at that moment,
they ranked somebody else top) whereupon X will be eliminated not Y
despite 99% preferring X.

That same voter may or may not
be able to influence later decisions by IRV;
and any given voter may in this way be able to influence many
such decisions, or few. Some IRV voters can thus
have tremendously more power than others.

By the way, proof of the 100% ignoring theorem
may be found in puzzle 34's answer:
http://rangevoting.org/PuzzlePage.html#p34

[Kevin Baas]

in the first sentence i clarified what i meant earlier when i used the term, which i acknowledged was not the common definition.

in the second sentence i mentioned another property with range voting more in line with the common definition.

in contrast, the entirety of your comment was offensive, derogatory, and not at all constructive.

[Kevin Baas]

so what you're saying is that at any given point in the tallying algorithm (not counting exhausting ballots), exactly one of the voter's ranking on each given ballot is considered.

and if no candidate reaches quota, then ballots that would otherwise have no effect on the outcome are transferred to the next choice, thus making their vote have a greater effect on the outcome.

and furthermore, in a multi-winner election, excess votes beyond that which are needed are transfered so that they can be re-used until each person's ballot counts for a single vote (as opposed to a fractional vote if its in excess).

conversely, this transferring of vote that allows a voter to maintain more voting power is not possible with range or approval voting.

[Warren D Smith]

With range or approval voting the algorithm never ignores any of the input,
and it treats all that input simply and fairly, i.e symmetrically.

But IRV ignores a lot of the input; and the subset of the input it uses,
can create the impression that some voters have much more power than others.
The powerful voter subset is not merely a random subset;
they in some elections can seem to be
selected in a very biased manner.

[Toby Pereira]

On Monday, 18 July 2016 16:06:19 UTC+1, Kevin Baas wrote:

 

So a person who votes for a compettive candidate, their vote counts a lot more than someone who voted for either a landslide winner or hell's snowball.

This is ideally what you want, since some people prefer X to Y by a small amount, whereas others prefer X to Y by a large amount. 

No, you want everyone's vote to count as much the same as possible,

I do get this position. Although some people prefer X to Y by a small amount and some by a large amount, when people vote strategically, it becomes less relevant. If a voter votes strategically by voting approval style by only giving the minimum or maximum score (not bullet voting, Kevin!), they basically have all the candidates in just two layers. It can then be argued that they haven't managed to make their vote count properly if they didn't approve one of the top two finishers and not approve the other. A strategic voter would generally attempt to do this regardless of how much they preferred one to the other.

With Condorcet voting, on the other hand, because every pair of candidates is compared head-to-head, you're getting a "full vote" in every head-to-head election and not having to decide which head-to-heads are worth participating in.

This is not to say that I prefer Condorcet voting to score or approval voting, because there are lots of other things to say as well, but this argument works on a very intuitive level.

[Toby Pereira]

I thought this rang a bell actually when I was typing it, and Warren has a page talking about "full support" in Condorcet methods - http://rangevoting.org/OssiDelusion.html - Warren says he's "not precisely sure what "fully supporting A versus B" means [to Mike Ossipoff]" but I'm pretty sure he meant the same thing I was talking about, which is what makes Condorcet methods intuitively appealing.

[Kevin Baas]

(This is not a reply to a particular comment, just a though.)

To make discussion easier we should probably make clear whether we are taking about single or multi winner. In single winner there's no need to transfer surplus votes because there's no second election to transfer then to. Though there may be a value in transferring deficit votes.

For single winner elections I think there is a value in either using allocation voting or using score going and then normalizing all ballots to sum=1. That way you can transfer deficit votes and there'd be no advantage to rating all 10'a and zeros.

I think this would be better than just a raw plurality style - first past the post - system. It would enable people who prefer an unpopular candidate to vote honestly while still having the "power" of their vote, if not equal, at least closer to the power of a person who votes for a leading (competitive) candidate.

[Kevin Baas]

A strategic voter on a range voting ballot would firstly vote all 10's and 0's (and yes, I've been using the term "bullet voting" all wrong this whoe time! Is there not a phrase for all 10'a and 0's?), and then secondly chose the 10's based in how they feel about the candidates and their likelihood of winning.

For instance, if there were two likely winners, both of which he like but one he preferred, he'd vote 10-0.

But if there were 4 likely winners, two of which he liked, though he liked the first one more, and two that he hated, he should vote 10-10-0-0.