Overview of Canada with recommended voting systems

[Warren D Smith]

http://rangevoting.org/CanadaOverview.html


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

[Bruce Gilson]

On Tue, Nov 17, 2015 at 5:01 PM, Warren D Smith <warre...@gmail.com> wrote:

http://rangevoting.org/CanadaOverview.html

​Unfortunately, your anti-Israel biases show up in your characterization of Israel as an "apartheid pseudo-democracy."​

 

​An Arab in Israel has much more freedom than a non-Muslim in any of the neighboring countries!​

[Jameson Quinn]

So the main suggestion that I had, that I don't see reflected in your document, is the idea of giving voters the option of voting in another riding of their choice. If voters are constrained to vote in their own riding, the boundaries can be drawn (deliberately or accidentally) so that certain incumbent MPs are basically "voter proof" by virtue of their party and riding. If "voters choose their representatives, not vice versa", then an unpopular MP can be voted out, no matter where they are; the local voters can still vote for their preferred party, but in another riding.

The other suggestion I'd made, that you didn't include, is that of forcing transfers of the smallest remainders first. This is probably not that big a deal.

Jameson

http://rangevoting.org/CanadaOverview.html

--
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]

To Bruce, I do not like Israel's pseudo-democracy, although it quite
likely is true that the Egyptian and Syrian non-democracies are worse.
The whole mideast is a craphole, and every time they do look like they
are getting a thriving democracy -- Lebanon, Iran -- then somebody destroys it,
namely Syria+Israel destroyed Lebanon's democracy, while USA+UK destroyed
Iran's. So if you want to know what countries are "the enemies of
democracy" now you know. Still looks to me like I'd prefer to be a
Jew living in Iran than an Arab in
the Gaza strip. But anyhow I would like to find a way to remove the
sentence about Israel if I can find a way to do so that still makes my point.
(Suggestions?)

To Jameson, I would like to enhance the files CanadaSAX.html
(some were by you) to include your & Toby's latest comments. Those files
are linked at the end but I have not updated them for a while. Even better:
you two update them.
The score-asset hybrid scheme I invented
appears in ScoreAssetHybrid.html and does do the least-juice-transfers-first
ploy in its step 7 (right?).

You & Toby did indeed have some interesting ideas, including the
"outside your riding" idea, which I indeed did not mention in my overview,
but they remain interesting ideas, and I'd like them to be available to those
who look into those files.

The main reason I did not talk about them was, I felt the ideas I did
present were sort of more logically central, and did not want reader
to get distracted by bells & whistles; and also I feel or hope II(C)1
is such a nice system that it obsoletes a lot of the ideas JQ and TP
and I were fooling with earlier. Do you agree? It really looks like
it makes things simple and good.

If you can blow a hole in it, please do so soon...

[William Waugh]

The variant of Range Voting (Score Voting) that you describe to them says to average the ratings a candidate receives from voters, allowing the voters to substitute "NO OPINION" for a rating, and not counting the NO OPINION entries in the average. I suppose that if write-ins are allowed and a voter does not write in a given write-in, such omission would be treated as another case of NO OPINION instead of full opposition. This raises the horror story that Adolph Hitler III stands for election, with the same policies as Adolph Hitler, and has 100 supporters, and no one else knows he is standing. He wins with 100% of his vote. If you propose a quota requirement or some fake ballots, I don't know how to justify the choice of value for the quota or the count of fake ballots. Therefore I suggest that the variant of Range Voting that should be proposed for Canada is the variant in which unmentioned candidates receive the lowest score.

William Waugh

[Clay Shentrup]

> one can adjust the #candidates, #voters, #issues,

I think you should say "number of", because my Twitter-tuned brained read that as "hashtag candidates" etc.

[Clay Shentrup]

>  Those systems' claims to be PR depend on "proportionality theorems" which say that if the voters behave in specified (usually rather extreme and "racist") ways, then the system guarantees electing a parliament whose "color" composition is the same as the electorate's

When you say "racist" and "color", I think a lot of readers may take that too literally. Could you use something else, like "factionalized"?

[Clay Shentrup]

On Tuesday, November 17, 2015 at 8:56:51 PM UTC-8, Clay Shentrup wrote:

>  Those systems' claims to be PR depend on "proportionality theorems" which say that if the voters behave in specified (usually rather extreme and "racist") ways, then the system guarantees electing a parliament whose "color" composition is the same as the electorate's

When you say "racist" and "color", I think a lot of readers may take that too literally. Could you use something else, like "factionalized"?

Sorry, I missed the part where you kind of explained that. Still, maybe better to avoid this term. 

[William Waugh]

The link 9 (nine) is dead.

On Tuesday, November 17, 2015 at 5:01:20 PM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

[Warren D Smith]

Canada does not allow write-ins.

[Clay Shentrup]

> Q = ∑1≤v≤V Simmons( ∑1≤j≤W [score awarded to winnerj by voter v on her ballot]/M ).

If this is, in your view, the best of your reform proposals, then I think you should explain it in plain English, to the best of your ability at least.

[Warren D Smith]

--The inner sum is voter v's opinion of the winner-committee as a summed score.
This is then transformed by the Simmons function and summed over all voters.
The Simmons function always increases, making it always better to please
any particular voter, but its rate of increase tapers off in just the right
way so that the more a voter is satisfied, the less influence she has
on Q. Hence, for example, a voter majority will of course exert the
most influence,
but after they are largely satisfied, an unsatisfied minority will
still have influence
and thus will be able to elect some winners of their own.

[Bruce Gilson]

On Tue, Nov 17, 2015 at 10:59 PM, Warren D Smith <warre...@gmail.com> wrote:

​[...]

 I would like to find a way to remove the
sentence about Israel if I can find a way to do so that still makes my point.
(Suggestions?)

​For:

The failed state of Germany, during the period between World Wars I and II, used this approach; and later it was adopted by Israel, now an apartheid pseudo-democracy. 

Write:

The failed state of Germany, during the period between World Wars I and II, used this approach; and later it was adopted by Israel, which has had its own crises in the past. ​

​Israel's biggest problem has been the hijacking of the system by one-issue parties (usually religious in nature) because neither of the two largest secular parties can form a government on its own, and so they sell out to the religious parties to gain enough votes for a coalition.​

 

[Warren D Smith]

Re mideast democracies, besides Lebanon & Iran
another excellent example was "the Kurds" who have had a functioning
democracy for decades. But the USA, because it is intent on
preventing democracy in the middle east,
has refused to recognize them as a "country." For all that time.
Instead, the USA set up a puppet pseudo-democracy in Iraq and declared
that it subsumed the Kurds. The Kurds, however, were not interested
in being conquered by a foreign power and having their democracy
terminated by the stroke of a pen, and largely ignored this
development.

Why does the USA, supposedly the champion of democracy, systematically
oppose, intentionally destroy, and/or refuse to recognize actual
democracies in the Middle East? While vigorously supporting some
extremely undemocratic regimes such as Saudi Arabia and the Shah of
Iran, and oppressive theocratic apartheid nuclear-rogue states like
Israel -- thus inspiring massive anti-USA hatred? Because, evidently,
the USA cares a lot more about other things than it cares about
democracy.

After having continued that policy for about 60 years, the USA has
indeed succeeded in what it tried to accomplish -- turning the Middle
East into a
hellhole. Of course, the USA is by no means entirely responsible for
that, but it systematically worked in that direction ever since the
end of world war II while lying about it the whole time.

If the USA actually were interested in championing democracy and making
their actions match their words, they'd
1. cut off aid to Israel.
2. treat Saudi Arabia the same way, or worse, as it treated Cuba.
3. support Iran, as a former democracy and current semi-democracy.
4. recognize the Kurds as a country.
If the USA cannot bring itself to do those things, then stay the hell
out of the Middle
East and stop making everything there worse.

[Warren D Smith]

> *​For:*
>
> *The failed state of Germany, during the period between World Wars I and

> II, used this approach; and later it was adopted by Israel, now an

> apartheid pseudo-democracy. *
>
> *Write:*
>
> *The failed state of Germany, during the period between World Wars I and

> II, used this approach; and later it was adopted by Israel, which has had

> its own crises in the past. ​*
>

--thanks: now reworded as

The failed state of Germany, during the period between World Wars I
and II, used this approach; and later it was adopted by Israel, which

also hasn't exactly been universally lauded. Those are two countries
Canada should not try to emulate. On the bright side, this approach
seems to have worked OK for the Netherlands. But Canada, as the
geographically second largest country with 240 times the land area of
Netherlands, probably would be insane to eliminate regionality.

[Warren D Smith]

JQ, I added a note about your to the end of
http://rangevoting.org/ScoreAssetHybrid.html

[Toby Pereira]

Your main reason for having five top-up MPs per region seems to be computability. If you think five would be the best or nearly-best number anyway then fine, but if computability wasn't an issue and five wasn't your preferred number under these conditions, then I would look at sequential election. Have you done any simulations/calculations on the effects of electing sequentially rather than all-at-once? Intuitively, I don't think the results would be massively different. So I'd pick my "favourite" number of top-up MPs and then look at the best way of electing them, rather than the other way round.

Also I'm not convinced by "optimum-psi" voting. It seems to be largely a non-sequential version of RRV. I've done calculations involving a non-sequential RRV myself and I didn't like the results. The way I did it, a voter's satisfaction score was the harmonic function of the number of candidates they had elected. (If they score an elected candidate 5/10 then this would count as having half a candidate elected.) This would be the Jefferson/D'Hondt version, but a Sainte-Lague version could equally be done. I know that the harmonic function is related to the digamma function that you use, so they are probably the same basic system.

First of all, it doesn't collapse to score voting in the single-winner version. You might see this as a feature rather than a bug, but I'm just putting it out there. For example:

1 voter: A=10, B=4

1 voter: A=0, B=4

A would win under score, but under non-sequential RRV, B would win.

A = HARMONIC [10/10] = 1

B = 2 * HARMONIC [4/10] = 1.03. http://www.wolframalpha.com/input/?i=2+*+HARMONIC+%5B4%2F10%5D

I've just looked at your Simmons graph and the figures seem to tally up exactly with the Sainte-Lague version of what I'm talking about. They are probably the same method. The example I've given is for D'Hondt, but the same principle still applies. It doesn't collapse to normal score voting for a single winner. But anyway, you might not see that as a problem.

The second problem, as I've detailed elsewhere, is that RRV (and therefore optimum-psi system) behaves horribly with score below the max. Read my post here: https://groups.google.com/d/msg/electionscience/BURkyIxZaBM/GdAcH2z7XkEJ

Basically while you get proportionality if everyone votes along party lines and gives max scores or zero to every candidate, this falls apart for scores lower than the max. An  example from tis post https://groups.google.com/d/msg/electionscience/BURkyIxZaBM/df0aQ9U0OlUJ :

4 voters: A=B=C=D=E=7; G=H=I=J=10; all other candidates = 0

1 voter: F=7; K=10; all other candidates = 0

If five candidates are elected, then it would be GHIJK. But then if more than five candidates are elected, it will move onto the candidates that received 7s, and these will then be elected in a less than proportional manner. Say there are 10 to be elected. Proportionally, it should be 8 to the larger faction and 2 to the smaller faction. If the first 9 elected are all the 10s and A to D, we would expect F to be preferred to E, proportionally speaking.

Using the 1/(1+SUM/MAX) ballot weight formula, the larger faction's ballots are weighted at 1/(1+68/10) = 0.128.

The smaller faction's ballots are weighted at 1/(1+10/10) = 0.5.

This gives E a score of 4*7*0.128 = 3.590.

F has a score of 1*7*0.5 = 3.5.

E is (disproportionally) elected.

This is using sequential RRV with D'Hondt divisors, but the principle is the same if you use Sainte-Lague and/or non-sequential RRV. This is easy to fix though, by converting non-max scores into approvals. As I detailed here https://groups.google.com/d/msg/electionscience/g_JwRuUvwak/HdLS5sYmCAAJ :

"For a maximum score of S, each voter is "split" into S parts, which we can name 1, 2, 3 ... S. If there is a voter called John, then effectively he becomes John1, John2... JohnS. If a voter awards a score of T (from 0 to S) to a candidate, then the first T parts of the voter approve the candidate. For example, if the maximum score is 10 and John gives a candidate a score of 5, then John1, John2, John3, John4 and John5 approve the candidate and John6, John7, John8, John9 and John10 do not."

This fix would also cause the method to collapse into normal score voting for a single winner.

At a very minimum I would suggest using this fix (or another like it), even if you ignore everything else.

Finally, the method fails what I call independence of commonly rated candidates. I know you have acknowledged this (and call it the Toby Pereira Criterion), but failure of this leads directly to more egregious failures of proportionality.

The example here - https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/nWMSo78whIQJ - is probably overly complex, but it demonstrates such a failure. A simpler but slightly less direct example is here - https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/hZUsOqQPS-EJ which I'll reproduce here:

"6 to elect

 

2 voters: U1, U2, A1
1 voter: U1, U2
3 voters: C1, C2, C3, C4

 

RRV seems to go for U1, U2, C1, C2, C3, C4, regardless of whether you use D'Hondt or Sainte-Laguë divisors.

 

I think the result should be U1, U2, A1, C1, C2, C3"

The fix for this would be to use an entirely different method, such as Ebert or Phragmen. These would ideally be used sequentially, however, because of likely Pareto/monotonicity failures with non-sequential elections.

[Toby Pereira]

On this, by adopting the fix, you also make the method a lot simpler. You would only be adding 1 + 1/3 + 1/5 + 1/7 etc. with nothing in between so there would be no need for the overly complex digamma function.

[Warren D Smith]

On 11/18/15, 'Toby Pereira' via The Center for Election Science

<electio...@googlegroups.com> wrote:
> Your main reason for having five top-up MPs per region seems to be
> computability. If you think five would be the best or nearly-best number
> anyway then fine, but if computability wasn't an issue and five wasn't your
>
> preferred number under these conditions,

--I definitely chose 5 because it was maximal subject to
computational feasibility. However, my choice (5,13), intuitively speaking
when you think about Canada's situation,
seems to be pretty good even if we had no computational limitations.

> then I would look at sequential
> election. Have you done any simulations/calculations on the effects of
> electing sequentially rather than all-at-once? Intuitively, I don't think
> the results would be massively different.

--they often might not be too different, but why rely on intuition and
hopes, when you can
just return the optimum pentad and avoid all possible aspersions?

> So I'd pick my "favourite" number
> of top-up MPs and then look at the best way of electing them, rather than
> the other way round.

--well, how would you choose that favorite?

> Also I'm not convinced by "optimum-psi" voting. It seems to be largely a
> non-sequential version of RRV. I've done calculations involving a
> non-sequential RRV myself and I didn't like the results. The way I did it,
> a voter's satisfaction score was the harmonic function of the number of
> candidates they had elected. (If they score an elected candidate 5/10 then
> this would count as having half a candidate elected.) This would be the
> Jefferson/D'Hondt version, but a Sainte-Lague version could equally be
> done. I know that the harmonic function is related to the digamma function
> that you use, so they are probably the same basic system.

--If psi voting with Delta=1/2 is done with APPROVAL style score voters
then I believe it would be equivalent to your system in Sainte-Lague version.
Psi voting with Delta=1 wiht approval-style voters ought to be
equivalent to your system d'Hondt version.
The reason is that C+Psi(x+1) = 1 + 1/2 + 1/3 + ... for x terms if
x is an integer, and only integers would arise if voters only vote
approval-style.
And C+Psi(x+1)/2 = 1 + 1/3 + 1/5 + 1/7 + ... for x terms if
x is an integer. These for appropriate constants C.
The Psi function is basically the nicest interpolation of these integer
results to real x.

> First of all, it doesn't collapse to score voting in the single-winner
> version. You might see this as a feature rather than a bug, but I'm just
> putting it out there.

--yes, I was aware of that.
I actually consider it a bug, but it is about the least-serious such a bug
could hope to be, because psi voting in the single winner case actually is
ISOMORPHIC to score voting. That is, if all voters regard the score scale
(which we'll regard as the real interval [0,1]) as somewhat
distorted, e.g. because
they all are wearing distorting-spectacles, with the distortion-map
arising from the functional inverse of the psi(Delta+x) function, then
when everybody puts ont he spectacles the psi voting system exactly equivalent
to score voting.

>For example:
>
> 1 voter: A=10, B=4
> 1 voter: A=0, B=4
>
> A would win under score, but under non-sequential RRV, B would win.
>
> A = HARMONIC [10/10] = 1
> B = 2 * HARMONIC [4/10] = 1.03.
> http://www.wolframalpha.com/input/?i=2+*+HARMONIC+%5B4%2F10%5D

--yes. The psi system is designed to obey a PR proportionality theorem
but only for "maximally racist" voters (e.g. approval-style voters)
based on the score-scale
[0,1]. I.e. there is a specific score-scale-width, namely 1 here,
"hardwired in."
If you switch to a smaller width, such as 0.7 which you had in some
googlegroup page you cited, then the PR theorem does not work.

Also, a related issue about PR systems is this. You might intuitively say:
"if all voters reverse their orderings (if the system uses rank-order
ballots) or
apply the reflection map x --> 1-x (if we use ratings ballots using score-scle
being the real interval [0,1]), then I want my multiwinner voting
system to "elect"
the same set of losers, as arose in the original election which was
trying to elect winners."
Kind of a reversal symmetry. Well, this intuition is totally wrong
and systems obeying this property are both impossible and undesirable.

> I've just looked at your Simmons graph and the figures seem to tally up
> exactly with the Sainte-Lague version of what I'm talking about. They are
> probably the same method.

--yes, see remarks above re approval-style and integer arguments.

> The example I've given is for D'Hondt, but the
> same principle still applies. It doesn't collapse to normal score voting
> for a single winner. But anyway, you might not see that as a problem.
>
> The second problem, as I've detailed elsewhere, is that RRV (and therefore
> optimum-psi system) behaves horribly with score below the max. Read my post
>
> here:
> https://groups.google.com/d/msg/electionscience/BURkyIxZaBM/GdAcH2z7XkEJ
>
> Basically while you get proportionality if everyone votes along party lines
>
> and gives max scores or zero to every candidate, this falls apart for
> scores lower than the max. An example from tis post
> https://groups.google.com/d/msg/electionscience/BURkyIxZaBM/df0aQ9U0OlUJ :
>
> 4 voters: A=B=C=D=E=7; G=H=I=J=10; all other candidates = 0
> 1 voter: F=7; K=10; all other candidates = 0
>
> If five candidates are elected, then it would be GHIJK. But then if more
> than five candidates are elected, it will move onto the candidates that
> received 7s, and these will then be elected in a less than proportional
> manner. Say there are 10 to be elected. Proportionally, it should be 8 to
> the larger faction and 2 to the smaller faction. If the first 9 elected are
>
> all the 10s and A to D, we would expect F to be preferred to E,
> proportionally speaking.

--yes, as I said, this problem arises since the psi functon is
designed for a fixed score-scale width, e.g. 1, and you switched to
width 0.7, and then PR
guarantees no longer happening.

> Using the 1/(1+SUM/MAX) ballot weight formula, the larger faction's ballots
>
> are weighted at 1/(1+68/10) = 0.128.
> The smaller faction's ballots are weighted at 1/(1+10/10) = 0.5.
>
> This gives E a score of 4*7*0.128 = 3.590.
> F has a score of 1*7*0.5 = 3.5.
>
> E is (disproportionally) elected.
>
> This is using sequential RRV with D'Hondt divisors, but the principle is
> the same if you use Sainte-Lague and/or non-sequential RRV. This is easy to
>
> fix though, by converting non-max scores into approvals. As I detailed here
>
> https://groups.google.com/d/msg/electionscience/g_JwRuUvwak/HdLS5sYmCAAJ :
>
> "For a maximum score of S, each voter is "split" into S parts, which we can
>
> name 1, 2, 3 ... S. If there is a voter called John, then effectively he
> becomes John1, John2... JohnS. If a voter awards a score of T (from 0 to
> S) to a candidate, then the first T parts of the voter approve the
> candidate. For example, if the maximum score is 10 and John gives a
> candidate a score of 5, then John1, John2, John3, John4 and John5 approve
> the candidate and John6, John7, John8, John9 and John10 do not."
>
> This fix would also cause the method to collapse into normal score voting
> for a single winner.

--well, I liked everything you said before, but I'm unsure whether to feel good
about this "fix."
It seems artificial and (what you yourself called) "unsymmetric."
I mean, should a voter who says "Joe=7" really be reckoned for PR
purposes as the same thing as 0.7 voters who say "Joe=10" and 0.3
voters who say "Joe=0"? And even if we accepted that, it gets
substantially more dubious if we declare that a
"Joe=8 Cal=4 Bob=3" voter is equivalent to
0.3 voters "Joe=10, Cal=10, Bob=10"
(which incidentally is a vote nobody sensible would ever cast)
0.1 voters "Joe=10, Cal=10, Bob=0"
0.4 voters "Joe=10, Cal=0, Bob=0".
ZERO voters "Joe=0, ..."
ZERO voters "Joe=10, Cal=0, Bob=10"

Hmmm. Well, first of all, the "unsymmetric" split-up can actually be
defended by
defining it in a symmetric manner, for example as follows:
"Among all possible ways to split an N-candidate score voter into a linear
combination of the 2^N approval-style votes, pick the unique one which
minimizes the sum of the squares of the coefficients." Or "pick the
unique one without any negative coefficients." Either ploy works.

Also, as you pointed out, if this is done, then we do not need the full power
of the psi(x) function for real arguments anymore, just the harmonic
sum 1+1/2+1/3+...
in the d'Hondt or the odd-harmonic sum 1+1/3+1/5+1/7+... in the
Sainte-Lague version,
with x terms, is all we need. Which makes life a bit simpler. The
splitup into 2^N
voters, unfortunately makes life more complicated, V voters turn into
V*2^N voters, horrible expansion.
But then when you MERGE voter-fragments all casting the same approval
ballot into "one voter" this reduces back down now to 2^N voters,
which might even be smaller than V.
Depends on the values of V and N, but in favorable cases you actually save work.
(This merge could be done elegantly algorithmically too, using the
wonderful world of binary.)

Unfortunately the numbers for Canada are not favorable?
Right now 5-7 candidates average run for each riding's MP seat.
So with my 18=13+5 plan,
we would expect 18*5=90 to 18*7=126 candidates running in a region.
2^90 is simply too damn large. But the saving grace is, if we only allow
voters to score candidates within their own riding, not the entire
13-riding region,
then only 2^5 to 2^7 approval-vote-types in each riding. For some
extreme riding
maybe 14 candidates, but 2^14 still is no problem.
OK, it could be done.

> At a very minimum I would suggest using this fix (or another like it), even
> if you ignore everything else.

--Hmm. Well, I'd have to say this is a very interesting, and
possibly very good, suggestion. I'm not sure whether to prefer it
over my original plan. It depends whether we think we should care
about "proportionality" for the voters who score Joe 7.
The meaning of proportionality was never clear for such voters. So maybe we
should just be happy and say that since it assures PR for extreme voters only,
therefore with non-extreme voters, we declare whatever it says, to be good.

But on the plus side, your way actually makes the computation easier
(Canada numbers above) and it does provide some PR notion you seem to like
re the Joe=7 voters.

So, I'd have to say your arguments have impressed me and I'm leaning
toward revising everything to do it your way. Plus, maybe this means
you should be a co-author, or anyhow get a lot of credit, for pointing
this out.

But this is annoying since it makes everything harder to explain+justify.
It's also bothering because, although your particular election
examples made your
particular notions of proportionality feel good, maybe there are other examples
where they will not feel so good. How do we know? How do we reckon all
examples, not just the cherrypicked ones?

The real way to compare is to quantify PRness using "2-stage bayesian regret"
which is something nobody ever did yet, though I know how.

> Finally, the method fails what I call independence of commonly rated
> candidates. I know you have acknowledged this (and call it the Toby Pereira
> Criterion), but failure of this leads directly to more egregious failures
> of proportionality.
>
> The example here -
> https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/nWMSo78whIQJ -
> is probably overly complex, but it demonstrates such a failure. A simpler
> but slightly less direct example is here -
> https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/hZUsOqQPS-EJ
> which
> I'll reproduce here:
>
> "6 to elect
>
> 2 voters: U1, U2, A1
> 1 voter: U1, U2
> 3 voters: C1, C2, C3, C4
>
> RRV seems to go for U1, U2, C1, C2, C3, C4, regardless of whether you use
> D'Hondt or Sainte-Laguë divisors.
>
> I think the result should be U1, U2, A1, C1, C2, C3"
>
> The fix for this would be to use an entirely different method, such as
> Ebert or Phragmen. These would ideally be used sequentially,
> however, because of likely Pareto/monotonicity failures with non-sequential
> elections.

--I'm unsure which winner set I like better in this case.

--Warren D. Smith

[Warren D Smith]

CORRECTION:

"And C+Psi(x+1)/2 = 1 + 1/3 + 1/5 + 1/7 + ... for x terms if
x is an integer."

should have read

And C+Psi(x+1/2)/2 = 1 + 1/3 + 1/5 + 1/7 + ... for x terms if
x>=0 is an integer.

--

[Warren D Smith]

CORRECTION:

"Among all possible ways to split an N-candidate score voter into a linear
combination of the 2^N approval-style votes, pick the unique one which
minimizes the sum of the squares of the coefficients." Or "pick the
unique one without any negative coefficients." Either ploy works.

should have said... actually only the first ploy works.
A counterexample would be 2*(1,1,1) = (1,1,0)+(0,1,1)+(1,0,1).

[Warren D Smith]

CLARIFICATION:

"But then when you MERGE voter-fragments all casting the same approval
ballot into "one voter" this reduces back down now to 2^N voters,"

should have said

But then when you MERGE voter-fragments all casting the same approval

ballot into "one voter" this reduces back down now to 2^N (weighted) voters

[Warren D Smith]

If we use the scheme suggested by Toby Pereira (and more recently above by me)
of regarding a score-style ballot as split into a bunch of weighted
approval-style ballots...
then what do we do about voters who intentionally left some candidate unscored?

I had regarded them as giving that candidate whatever his average
score (among the voters who *did* score him) was. We can still do
that with the split-up plan, but it's nastier.
Instead of 2^C possible approval-style ballots with C candidates,
there are now 3^C because "neither approve nor disapprove" is a 3rd
choice too. And then we'd have to optimize psi quality score really
using psi function -- we could not just say "only harmonic sums are
needed so we can dodge the need for psi function."

The 3^C thing is quite a severe hit.

[Toby Pereira]

I was just looking at your proposals again and your II(C)1 seems a little strange. If I've understood this correctly, voters only score the candidates in their own riding, so the final five seats are a competition between candidates who are essentially standing in entirely separate elections. I'm not sure what would result from this, but presumably you'd need the populations of these ridings to be pretty much exactly the same. Also the result would look unbalanced in that some ridings have more MPs than others without this being due to support from outside their riding. I'm pretty sure I've understood it correctly, given this: "As far as each voter is concerned, she simply provides a score (on an 0-9 numerical scale, if M=9) for each candidate in her riding, i.e. each ballot is no more and no less complex than scheme I(A). "

On Thursday, 19 November 2015 14:58:05 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 11/18/15, 'Toby Pereira' via The Center for Election Science
<electio...@googlegroups.com> wrote:
> Your main reason for having five top-up MPs per region seems to be
> computability. If you think five would be the best or nearly-best number
> anyway then fine, but if computability wasn't an issue and five wasn't your
>
> preferred number under these conditions,

--I definitely chose 5 because it was maximal subject to
computational feasibility.  However, my choice (5,13), intuitively speaking
when you think about Canada's situation,
seems to be pretty good even if we had no computational limitations.

> then I would look at sequential
> election. Have you done any simulations/calculations on the effects of
> electing sequentially rather than all-at-once? Intuitively, I don't think
> the results would be massively different.

--they often might not be too different, but why rely on intuition and
hopes, when you can
just return the optimum pentad and avoid all possible aspersions?

> So I'd pick my "favourite" number
> of top-up MPs and then look at the best way of electing them, rather than
> the other way round.

--well, how would you choose that favorite?

I suppose I'd look at the proportion/number of MPs that I'd want to be local representatives, and how proportional the overall result would become for a given number of top-up MPs and what I'd regard as acceptable for this. Also, while six or more top-up MPs might require sequential election and thus give less proportionality than non-sequential election, this is overcome by the fact that having more top-up MPs makes the result more proportional in itself. Note that this isn't a criticism of having five top-up MPs, but just questioning the reason that you chose five.

 

> Finally, the method fails what I call independence of commonly rated
> candidates. I know you have acknowledged this (and call it the Toby Pereira
> Criterion), but failure of this leads directly to more egregious failures
> of proportionality.
>
> The example here -
> https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/nWMSo78whIQJ -
> is probably overly complex, but it demonstrates such a failure. A simpler
> but slightly less direct example is here -
> https://groups.google.com/d/msg/electionscience/f72vTfY0sC8/hZUsOqQPS-EJ
> which
> I'll reproduce here:
>
> "6 to elect
>
> 2 voters: U1, U2, A1
> 1 voter: U1, U2
> 3 voters: C1, C2, C3, C4
>
> RRV seems to go for U1, U2, C1, C2, C3, C4, regardless of whether you use
> D'Hondt or Sainte-Laguë divisors.
>
> I think the result should be U1, U2, A1, C1, C2, C3"
>
> The fix for this would be to use an entirely different method, such as
> Ebert or Phragmen. These would ideally be used sequentially,
> however, because of likely Pareto/monotonicity failures with non-sequential
> elections.

--I'm unsure which winner set I like better in this case.

Well that was a slightly watered down version to keep it simple, but I also linked to a more clear-cut example which I deemed too long to reproduce in that post. But I might do that anyway now. The take home message from it is that factions who partially agree (as opposed to fully agreeing or not at all) lose out as a result of this. I'll quote directly from the post https://groups.google.com/forum/#!msg/electionscience/f72vTfY0sC8/nWMSo78whIQJ :

"I've been thinking more about the possible implications of failing my "independence of commonly rated candidates" criterion. ("Commonly" might be better than "universally" because I don't want it to imply that every voter has to rate them equally because it should apply for subfactions within factions etc).

 

To recap, the example I've been using is:

 

6 to elect, approval voting (or score voting where voters just top rate or bottom rate)

 

20 voters: A, B, C, D, E, F
10 voters: A, B, C, G, H, I

 

There may be some debate whether to elect A, B, C, D, E, F or A, B, C, D, E, G.

 

But what if we add two further factions that vote independently of anyone else?

 

Faction 1a (20 voters): U1, U2, U3, U4, U5, U6, A1, A2, A3.....

Faction 1b (10 voters): U1, U2, U3, U4, U5, U6, B1, B2, A3.....

Faction 2a (20 voters): C1, C2, C3.....

Faction 2b (10 voters): D1, D2, D3.....

 

I've named them 2a and 2b because even though they have no candidates in common, I'm going to assume that the candidates they vote for all happen to be from the same party. This is important. Also, I've left it open-ended how many candidates each faction votes for, so we can fill any number of seats.

 

Let's say there are 48 seats (sorry for the high number but I struggled to get this to work). What's the best result? I would argue that factions 1a and 1b should get 24 candidates between them, as should 2a and 2b because that gives party proportionality for one thing. For me, the ideal result would be:

 

U1-U6, A1-A12, B1-B6, C1-C16, D1-D8

 

This means that in terms of number of seats, we have:

 

Faction 1a: 18

Faction 1b: 12

Faction 2a: 16

Faction 2b: 8

 

Perhaps, if you argue against my criterion, you might have:

 

U1-U6, A1-A14, B1-B4, C1-C16, D1-D8

 

This gives:

 

Faction 1a: 20

Faction 1b: 10

Faction 2a: 16

Faction 2b: 8

 

However, according to my calculations, RRV (and I think probably several other methods) would give:

 

U1-U6, A1-A12, B1-B3, C1-C18, D1-D9

 

This gives:

 

Faction 1a: 18

Faction 1b: 9

Faction 2a: 18

Faction 2b: 9

 

This means that the party that 1a/1b have in common have 21 candidates elected, and 2a/2b have 27 candidates elected. This is arguably not just an undesirable result but a proper violation of proportionality! It might superficially look more proportional, but I don't see it like that, and factions 2a/2b have gained candidates from the co-operation of 1a/1b."

[Toby Pereira]

On Thursday, 19 November 2015 14:58:05 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Hmmm.  Well, first of all, the "unsymmetric" split-up can actually be
defended by
defining it in a symmetric manner, for example as follows:
"Among all possible ways to split an N-candidate score voter into a linear
combination of the 2^N approval-style votes, pick the unique one which
minimizes the sum of the squares of the coefficients."  Or "pick the
unique one without any negative coefficients."  Either ploy works.

Would you be able to explain that for dummies? The way I did it, each voter would just be split into [max score] parts, so the maximum number of approval ballots would be [voters] * [max score]. 

[Warren D Smith]

On 11/19/15, 'Toby Pereira' via The Center for Election Science

<electio...@googlegroups.com> wrote:
> I was just looking at your proposals again and your II(C)1 seems a little
> strange. If I've understood this correctly, voters only score the
> candidates in their own riding, so the final five seats are a competition
> between candidates who are essentially standing in entirely separate
> elections.

--true. My thinking had been... There would be 2 ways to go.
(i) We could have the entire 13-riding region vote on
all the candidates in that region, which would be about 100 candidates,
which seems too much to ask voters to do.
(ii) Or we could have each riding only score its own candidates, i.e. about 8.
Now IF we assume that (a) each riding knows more about its own candidates then
the ones in foreign ridings, and (b) if everybody knew everything
then the scores awarded to candidate X from riding Y would not depend
much on Y,
then we might as well save work and just get the scores about X from his
home-riding's voters only, since their average will be the
same under assumption b as if all ridings had scored him, and we get more
knowledge this way due to assumption a, which may well more than compensate for
a somewhat flawed assumption b.

And the fact all 13 ridings come from the same region of Canada also helps
make b true.

> I'm not sure what would result from this, but presumably you'd
> need the populations of these ridings to be pretty much exactly the same.

--yes, if they differed enough to invalidate assumption b, that'd be a problem.
But I'm hoping the benefits of my approach outweigh the cost of this problem.

> Also the result would look unbalanced in that some ridings have more MPs
> than others without this being due to support from outside their riding.

--a beneficial side effect of that is that it makes MPs want their home ridings
to have high voter turnout. As opposed to, right now in the USA,
where republicans are trying hard to DISCOURAGE voter turnout by
passing laws which make it harder to vote.
Point is, this would prevent/solve the USA's present pathology.

>> --well, how would you choose that favorite?
>
> I suppose I'd look at the proportion/number of MPs that I'd want to be
> local representatives, and how proportional the overall result would become
> for a given number of top-up MPs and what I'd regard as acceptable for
> this.

--yah, well, I did that. My proposal has regions each made of 13 ridings, and
Canada-wide at present there would be 17 regions in all for 17*18 MPs
in all. This is a smaller number of larger regions than in all three
FairVote Canada proposals, hence obtains more accurate PR than they
can. Each region having 18 MPs means parties
with 1/19 support will win seats, which is pretty decent PR, a good deal better
than all 3 FairVote Canada proposals. Which incidentally are described here:
http://rangevoting.org/CanadaVoteHistory.html

> Also, while six or more top-up MPs might require sequential election
> and thus give less proportionality than non-sequential election, this is
> overcome by the fact that having more top-up MPs makes the result more
> proportional in itself. Note that this isn't a criticism of having five
> top-up MPs, but just questioning the reason that you chose five.

--that is true, but I don't think the extra PR accuracy you'd get from
>=6 is worth it.
(And evidently neither does FairVote Canada.) Obtaining PR accuracy of 0.01%
would be silly because there would be plenty of effects that'd dwarf the 0.01%.
My point is, beyond some level it becomes silly. And there are
benefits to regionality,
i.e. keeping the size of regions small, such as what I was calling the
"rain effect,"
which would be lost if you went to larger regions in a silly quest to
get more PR accuracy. The point is, those benefits probably outweigh
the extra-PR-accuracy benefits.

So the basis for thinking it is somewhat intuitive, but I think my
numbers were near-optimal choices.

--I'm not sure what is the moral from this.
Look, my current thinking, as influenced by you (TP) is that
the "split up into weighted approval ballots" trick
makes you largely satisfied about your bad-PR election examples?

If so, that's good, because it now seems there are nice algorithmic and
nice definition ways to get election methods equivalent to doing that trick,
but without subjecting yourself to a lot of pain (even though one
might naively have thought pain was needed).

If not, i.e. you are still unsatisfied even with the split-up trick
put in, then not so good,
and then the question is what to do about it.

--

[Jameson Quinn]

My objection is still that this allows "voter-proof" MPs. Consider a riding that is 50% party A, 30% party B, 20% C and other. Say the incumbent party A candidate, Susie Slimeball, is involved in a minor corruption scandal; enough to tarnish her image, but not enough to force her to resign. She still has enough pull in the party — or perhaps just sheer luck — to make sure that no other credible party A candidate runs in the riding.

Now the party-A voters in that riding have an ugly choice. They can vote for Slimeball, ensuring that party A gets their support in terms of overall seats; or they can give her an honest bad rating, but see party A also suffer.

Of course, you can always say "that shouldn't happen". The party "should" avoid having slimeballs in the first place; when such people arise, it "should" be able to find other credible candidates to run against them; and a litany of other "shoulds". But "should" has never been a very strong word in politics.

This is even worse when you think of gerrymandering: Susie could make sure her district is explicitly drawn to favor her. Or instead of Susie Slimeball, it could just be Frank Fringe; somebody who does not represent their party base. In all of these cases, the most likely result is that a candidate is elected, even if some other losing candidate from party A in a less-favorable district is actually Pareto superior (favored by all voters over Susie/Frank, including both voters in party A and those in other parties).

This is the purpose of my "allow voters to opt in to other ridings" idea. It can be simply (and honestly) promoted as "voters should choose their representatives, not vice versa". And it allows the party A voters in Susie's district to shop around for a party A candidate in another district whom they can support more honestly. 

As a technical detail: if you're using this system, you should probably use totals and not averages for comparison across districts. Votes from other areas will probably not rate as many candidates above bottom, so they will basically function as boosts to a single candidate, not full ratings of all the candidates.

I think this is an important change.

I also think your page is too complex, and clearly needs an "executive summary" which DOESN'T EVEN ENUMERATE ALL THE OPTIONS, just says "there are various options and we think this one is best."

[Warren D Smith]

I had sought maximum simplicity by making only a single "score the
canddts in your riding" ballot. But does this, as JQ points out,
cause a nasty issue with "Susie Slimeball"?
Not really. Remember with my plan there are no party labels and the
PR is candidate-based. So by honestly rating Susie Slimeballl as
"crap," you as a voter are NOT hurting Susie's party, EXCEPT for in
her seat alone. Which, I think, is exactly the way it should be,
that'll motivate her party to improve in that riding.

Gerrymandering is not a problem in Canada now, due to their excellent
independent
line-drawing body Elections Canada. If it isn't a problem now using
FPTP in ridings,
it will be less of a problem with the PR plan in question. Unless
future Elections Canada
gets corrupted.

[Toby Pereira]

I was thinking more about this. Because the ridings have their own separate elections but are artificially forced into a proportional top-up election, the election of a candidate in riding A would have no effect on any voting weights of the voters in other ridings. So it should actually make no difference whatsoever if you elect sequentially or all-at-once. The five top-up seats are likely to go to the five candidates who were closest to winning the riding election - with the possible exception of a few "clonelike" candidates. The proportional part of the election looks like a red herring. It's simply that 13 ridings get an MP each, and five lucky ridings also get another one (normally if their election was close). I suppose it's theoretically possible for one riding to get three MPs, but I don't think I can see this ever happening in practice.

Jameson wants voters to be able to use a ballot for a different riding instead. Whether it's this or something else, I think there has to be some sort of "cross-voting" allowed to avoid the accusation that you've just got separate ridings with some getting two MPs and most getting one. There has to be a sense in which this is "one election" rather than several independent elections but with the results being put together in an unsatisfactory way. I understand your point that having all candidate names from the region on the ballot would be too much, and I would agree with it, but there are still other possibilities, including Jameson's idea and a limited write-in facility. Also all the ideas previously discussed involving delegated votes could be used at regional level even if they were thought up with national level in mind.

 

--I'm not sure what is the moral from this.
Look, my current thinking, as influenced by you (TP) is that
the "split up into weighted approval ballots" trick
makes you largely satisfied about your bad-PR election examples?

If so, that's good, because it now seems there are nice algorithmic and
nice definition ways to get election methods equivalent to doing that trick,
but without subjecting yourself to a lot of pain (even though one
might naively have thought pain was needed).

If not, i.e. you are still unsatisfied even with the split-up trick
put in, then not so good,
and then the question is what to do about it.

Well, there are two distinct problems that I've identified. One is that the method you originally proposed doesn't work so well for scores other than the max score. Turning scores into approvals solves that. The other separate problem is the failure of "independence of commonly rated candidates". Turning scores into approvals does not solve that. However, using Phragmen or Ebert does, even they potentially introduce other problems. But as I mentioned in the other discussion thread (Ebert proportionality) * (proprtion of overall support) could be a good compromise if it's computationally viable. Other than that, it just depends how big a problem failing independence of commonly rated candidates is deemed to be.

[Warren D Smith]

Ebert and Phragmen seem to be too computationally hard, and also not
simple enough to describe easily enough. (And by the way, Ebert never
proved he got PR -- I think he does, but I did not see anybody prove
it.)
So unless we can somehow make simpler-to-describe versions of their
ideas which also run faster, they need to be abandoned for Canada
purposes.
(I tried, but did not succeed well enough.)

Re "independent elections in different ridings, with some lucky ridings..."
well, that is pretty much a correct view of it, but I do not see why
that is a bad thing.
A riding which thinks it has 2 good candidates, gets two MPs.
What's wrong with that? That's the behavior we want.

If we instead had that riding #1 expresses opinions about MP
candidates in ridings #2,3,...,13 and in riding #1 about 1,3,4,...,13
and ... and then the best get to be top-up MPs, what the hell more
have we accomplished -- we just made the ballots huge with 100
candidates , and the "lucky" MPs are still the ones that tend to get
liked a lot, just like before, only now it's liking by people who do
not know as much about them.

I mean, I could do all that, and then it would not be "independent
elections" anymore, it'd
all be one entangled mess, but it seems to me the advantages got by
that would be small and would be outweighed by the disadvantages of
bigger ballots and less-knowledgable voters.

[Warren D Smith]

> I mean, I could do all that, and then it would not be "independent
> elections" anymore, it'd
> all be one entangled mess, but it seems to me the advantages got by
> that would be small and would be outweighed by the disadvantages of
> bigger ballots and less-knowledgable voters.

--and another disadvantage would be, that'd make candidates campaign
in the whole region not just their riding, increasing expenses by
factor 13,
hence increasing the pernicious influence of money.

Why?

[Warren D Smith]

I actually do like the whole asset voting, or "delegated votes" idea, and used
it in II(C)2, but it has been criticized as too hard to describe, and
also involving somewhat suspicious or suspicion-generating
negotiations or transfer procedures, which make it hard to "sell" the
method to the public... so that I was encouraged to eliminate
vote-delegation and make the voters alone control everything,
resulting in II(C)1.

There also are other versions of II(C)2 such as Jameson Quinn having
voters who can delegate to somebody in a foreign riding. Those might
be better, but again, we pay the price of extra complexity on the
ballot and in the method description, and extra goofy and somewhat
arbitrary rules need to be added.

It's really quite tough. You improve things in one area such as
"better delegation" whereupon they get worse in another department
such as "simplicity" and these are all pretty crucial.

-----

Re "independence of commonly rated candidates"
suppose there some universally-thought-good candidates AAA.
Then once AAA are elected, the alleged problem is, that
the remaining seats will not be proportionally allocated by "color."
If there were no AAA, they would have been.

But how about this response: AAA being rated good by all,
can be regarded as already consisting of
all colors mixed in whatever proportions I feel like saying!
I'll just pick those color proportions inside AAA
in such a way that the total parliament,
which you just criticized as not being color proportional, actually IS
color-proportional. So be happy?

Well, I'm not sure I like that response... there really are differences
between Phragmen and Thiele... but I guess it is a defensible response.

[William Waugh]

How about use separate sections of ballot for the local election and for the top-ups? Wouldn't that simplify the reasoning necessary to understand the properties and likely effects of any proposed system?

[Warren D Smith]

How about use separate sections of ballot for the local election and
for the top-ups? Wouldn't that simplify the reasoning necessary to
understand the properties and likely effects of any proposed system?

--sure, but now your ballot is more complicated. It now has 2 parts
instead of 1, and the 2nd part is MUCH larger.

[Toby Pereira]

On Friday, 20 November 2015 18:37:26 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

Ebert and Phragmen seem to be too computationally hard, and also not
simple enough to describe easily enough.  (And by the way, Ebert never
proved he got PR -- I think he does, but I did not see anybody prove
it.)
So unless we can somehow make simpler-to-describe versions of their
ideas which also run faster, they need to be abandoned for Canada
purposes.
(I tried, but did not succeed well enough.)

It's quite easy to prove the proportionality of Ebert. I can do it now in an informal manner. We will assume perfect party voting (or racist voting as Warren might call it). Perfect proportionality according to the Ebert method is when the sum of the squares of the voters' loads is the lowest it can possibly be, which is when they are all equal. If you have two factions and faction A is k times the size of faction B, then this method is properly proportional if Ebert would award faction B x candidates and faction A kx candidates. So let's say faction A has kb voters, and faction B has b voters. If faction B has x candidates elected, then the load per voter is x/b (each elected candidate providing 1 unit of load spread among its voters). If faction A has kx candidates elected, then the load per voter is kx/kb = x/b. So minimising the squares of the voters' loads happens when there is perfect party proportionality.

Phragmen and Ebert are analogous to D'Hondt and Sainte-Lague party list voting respectively and collapse into these methods if everyone votes along party lines. So if you prefer Sainte-Lague to D'Hondt (which you appear to), it's Ebert rather than Phragmen you should be looking at more seriously.

 

Re "independent elections in different ridings, with some lucky ridings..."
well, that is pretty much a correct view of it, but I do not see why
that is a bad thing.
A riding which thinks it has 2 good candidates, gets two MPs.
What's wrong with that?  That's the behavior we want.

If we instead had that riding #1 expresses opinions about MP
candidates in ridings #2,3,...,13 and in riding #1 about 1,3,4,...,13
and ... and then the best get to be top-up MPs, what the hell more
have we accomplished -- we just made the ballots huge with 100
candidates , and the "lucky" MPs are still the ones that tend to get
liked a lot, just like before, only now it's liking by people who do
not know as much about them.

I mean, I could do all that, and then it would not be "independent
elections" anymore, it'd
all be one entangled mess, but it seems to me the advantages got by
that would be small and would be outweighed by the disadvantages of
bigger ballots and less-knowledgable voters.


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

One other problem with this method is that it doesn't give a very high level of proportionality. The proportionality is locked within each riding, so a candidate cannot get elected with 1/18 or 1/19 of the support. If getting a third of the vote is what's required to get elected in an elect-two election, a candidate would need a third of the support in their riding just to have a chance of becoming one of the 5 out of 18 runners up that get elected. It makes no difference having enough support across the whole region because it's not really regionally proportional. Also, parties that keep just missing out don't save anything up as a result of it and so don't end up winning a seat.

Also I don't think you'd be able to sell it. People would just be saying that it's unfair that a neighbouring riding got two MPs because their election was closer. They would not see them as regional representatives because they were elected exclusively by voters from their own riding.

If ridings are kept separate in this manner, it would probably be better to have nine ridings per region, and elect two in each, but that's still not very proportional

But this is the problem with proportionality versus ballot simplicity. You've kept the simplicity but done away with almost all of the proportionality. If you had a country with STV (or some other PR system) that elected just two candidates per riding, it wouldn't be seen as very proportional. But this is less proportional than that because on average it's elect 1.38. Combining the regions for the proportional count does not give regional proportionality.

[Toby Pereira]

On Friday, 20 November 2015 19:06:39 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

I actually do like the whole asset voting, or "delegated votes" idea, and used
it in II(C)2, but it has been criticized as too hard to describe, and
also involving somewhat suspicious or suspicion-generating
negotiations or transfer procedures, which make it hard to "sell" the
method to the public... so that I was encouraged to eliminate
vote-delegation and make the voters alone control everything,
resulting in II(C)1.

There also are other versions of II(C)2 such as Jameson Quinn having
voters who can delegate to somebody in a foreign riding.  Those might
be better, but again, we pay the price of extra complexity on the
ballot and in the method description, and extra goofy and somewhat
arbitrary rules need to be added.

It's really quite tough.  You improve things in one area such as
"better delegation" whereupon they get worse in another department
such as "simplicity" and these are all pretty crucial.

If asset/delegation is unlikely I'm sure we can still do better than elect 1.38 (on average).

 

-----

Re "independence of commonly rated candidates"
suppose there some universally-thought-good candidates AAA.
Then once AAA are elected, the alleged problem is, that
the remaining seats will not be proportionally allocated by "color."
If there were no AAA, they would have been.

But how about this response: AAA being rated good by all,
can be regarded as already consisting of
all colors mixed in whatever proportions I feel like saying!
I'll just pick those color proportions inside AAA
in such a way that the total parliament,
which you just criticized as not being color proportional, actually IS
color-proportional.  So be happy?

Well you could say that, but this sort of works on the assumption that voters are assigned to specific candidates, whereas I prefer a sort of proportionality that takes voters' opinions of all candidates into account, which is one reason to use score rather than STV. And even that breaks down anyway with that convoluted example I gave when one group of voters had fewer representatives than another of equal size simply because they partially agreed with each other. If they fully agreed or not at all, they would have been properly represented. And I don't think any way of talking round the subject can turn that result into proportionality.

 

Well, I'm not sure I like that response... there really are differences
between Phragmen and Thiele...  but I guess it is a defensible response.

And don't forget Ebert! 

[Jameson Quinn]

2015-11-20 14:06 GMT-05:00 Warren D Smith <warre...@gmail.com>:

I actually do like the whole asset voting, or "delegated votes" idea, and used
it in II(C)2, but it has been criticized as too hard to describe, and
also involving somewhat suspicious or suspicion-generating
negotiations or transfer procedures, which make it hard to "sell" the
method to the public... so that I was encouraged to eliminate
vote-delegation and make the voters alone control everything,
resulting in II(C)1.

I think intraparty results should be based on ballots only; this prevents "party power brokers" from being elected in spite of the voters.

As for interparty transfers, I think asset is best. This can be either "free market" asset, or "bottom up elimination" asset. I weakly prefer the latter as giving more power to voters, but it's not really that important. (As for the question of whether either of these would be seen as "too complicated" by the voters: I actually think that both of these can be explained in equally-intuitive terms. Sure, full-blown asset with no explicit role for the parties is technically simpler than this; but in terms of explaining to voters what's going on and what guarantees the system is giving them, I think that this kind of "intraparty open-list, interparty asset" system is actually more intuitive.)

 

There also are other versions of II(C)2 such as Jameson Quinn having
voters who can delegate to somebody in a foreign riding.  Those might
be better, but again, we pay the price of extra complexity on the
ballot and in the method description, and extra goofy and somewhat
arbitrary rules need to be added.

You think "vote in any riding you want" is "goofy"? I disagree. It's certainly simple to say. Yes, it requires a couple of FAQs:

Q. Why let people vote in different ridings?

A. Voters should be able to choose their representative, not vice versa. If you generally support a given party, but have some specific problem with your local candidate for that party, you should be able to vote for a better candidate elsewhere. Furthermore, this lets you decide what riding represents you best. If you live just on one side of a border, but your job, school, and/or friends are mostly on the other side, you should be able to vote in the riding you actually identify with.

Q. Wouldn't this allow people from other ridings to "hijack" the outcome in mine?

It's true that if candidate A gets a lot of support locally, but candidate B gets even more support from "outside" voters, then B could win the local seat. But those votes for candidate A are not wasted. They still count for A's party, so, since A's support was strong, that party will almost certainly get at least of the "proportional" seats, and one of those seats will probably go to A. Even in the rare cases where this doesn't happen, either because A's party doesn't have enough extra votes to add up to a full extra seat, or because some other candidate from the same party got even more votes without winning locally, A will be first in line to decide which (if any) other party to allocate those votes to.

...But I think those FAQs are pretty clear, and that it's better to add this small amount of explanation of something that is not actually a problem, than to simplify and allow an actual problem to creep in.

...

As far as I can tell, in your classification, the system I'm advocating is II(c)2 with cross-riding votes, while you prefer II(c)1, that is, range locally and psi regionally. I think there are multiple problems with your proposal:

1. Too complicated. It may be simple and in fact beautiful from a mathematical perspective, but any time you have to talk about the "'psi function', which behaves much like the more commonly known 'logarithm' function", you have already lost 99% of voters. I'm a PhD student in a mathematical field, and I think most of my peers couldn't understand this proposal without at least 5 minutes and a pad of paper/blackboard. These people are all easily 99th percentile for the kind of math literacy that lets them read and understand something like this. 

2. By "simplifying" away parties, you are removing some of the most important proportional guarantees.

3. I think that with a partisan electorate you could still get the "Susie Slimeball" pathology.

4. I suspect that there are some strange effects of non-approval-style score ballots, which would make these strategically favored/disfavored in certain situations, leading to unequal voting power for strategically sophisticated or lucky groups.

 

It's really quite tough.  You improve things in one area such as
"better delegation" whereupon they get worse in another department
such as "simplicity" and these are all pretty crucial.

-----

Re "independence of commonly rated candidates"
suppose there some universally-thought-good candidates AAA.
Then once AAA are elected, the alleged problem is, that
the remaining seats will not be proportionally allocated by "color."
If there were no AAA, they would have been.

But how about this response: AAA being rated good by all,
can be regarded as already consisting of
all colors mixed in whatever proportions I feel like saying!
I'll just pick those color proportions inside AAA
in such a way that the total parliament,
which you just criticized as not being color proportional, actually IS
color-proportional.  So be happy?

Well, I'm not sure I like that response... there really are differences
between Phragmen and Thiele...  but I guess it is a defensible response.



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

[William Waugh]

How about use separate sections of ballot for the local election and
for the top-ups? Wouldn't that simplify the reasoning necessary to
understand the properties and likely effects of any proposed system?

--sure, but now your ballot is more complicated. It now has 2 parts
instead of 1, and the 2nd part is MUCH larger.

I have been skimming the discussion and if I understand right, some of your proposals would limit the set of candidates a voter would have to consider for the top-up phase to those running in a "region" of five ridings or so. So that would reduce the number of candidates the voter would have to consider.

Another way to limit the physical size of the top-up section of the ballot would be to leave off listing the candidates, but permit the voter to enter a code number, or a few. This would work for voters willing to research ahead of time to decide which candidacies to support, or voters who trust the advice of party operatives or groups advocating on policy. In some variants, a code number could represent a slate of candidates with scores rather than a single candidate.
 

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

William Waugh 

[Toby Pereira]

On Friday, 20 November 2015 22:59:21 UTC, Toby Pereira wrote:

On Friday, 20 November 2015 19:06:39 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

I actually do like the whole asset voting, or "delegated votes" idea, and used
it in II(C)2, but it has been criticized as too hard to describe, and
also involving somewhat suspicious or suspicion-generating
negotiations or transfer procedures, which make it hard to "sell" the
method to the public... so that I was encouraged to eliminate
vote-delegation and make the voters alone control everything,
resulting in II(C)1.

There also are other versions of II(C)2 such as Jameson Quinn having
voters who can delegate to somebody in a foreign riding.  Those might
be better, but again, we pay the price of extra complexity on the
ballot and in the method description, and extra goofy and somewhat
arbitrary rules need to be added.

It's really quite tough.  You improve things in one area such as
"better delegation" whereupon they get worse in another department
such as "simplicity" and these are all pretty crucial.

If asset/delegation is unlikely I'm sure we can still do better than elect 1.38 (on average).

One thing you could do is take a score for a candidate to be the score for that party (i.e. all candidates from that party) in the top-up phase. Alternatively there could be a separate box for "party score" next to each candidate. If left blank this would default to the score given to the candidate. The only problem with this is independents, which I don't think we can lump together as a "party" for these purposes. There could be a line on the ballot box under the list of candidates standing in the riding and a list of all independent candidates in the region, so voters could give them a score purely for the top-up phase. It would complicate the ballot, but what is the optimum simplicity/proportionality trade-off? And people could ignore this section. And how many independents would realistically stand in 13 ridings? In the UK in the 2015 election, I think it was about half an independent per constituency, so that would about mean six or seven here - not excessive. I'm not sure if Canada is different and also a better voting system might encourage more independents to stand. But it might have the opposite effect because the system might mean that they trust the parties more.

[Jameson Quinn]

2015-11-20 18:18 GMT-05:00 William Waugh <2knuw...@snkmail.com>:

How about use separate sections of ballot for the local election and
for the top-ups? Wouldn't that simplify the reasoning necessary to
understand the properties and likely effects of any proposed system?

--sure, but now your ballot is more complicated. It now has 2 parts
instead of 1, and the 2nd part is MUCH larger.

I have been skimming the discussion and if I understand right, some of your proposals would limit the set of candidates a voter would have to consider for the top-up phase to those running in a "region" of five ridings or so. So that would reduce the number of candidates the voter would have to consider.

Another way to limit the physical size of the top-up section of the ballot would be to leave off listing the candidates, but permit the voter to enter a code number, or a few. This would work for voters willing to research ahead of time to decide which candidacies to support, or voters who trust the advice of party operatives or groups advocating on policy. In some variants, a code number could represent a slate of candidates with scores rather than a single candidate.
 

I've made proposals based on both of these possibilities, since at least when I created PAL over 2 years ago. My current thinking is that it's simpler (and probably an easier sell) to just let voters opt to get a ballot from another riding.

(In an American context, where you have multi-page ballots including local initiatives and school board/sheriff/judge/clerk/dogcatcher, this "select your ballot" would be by race, which would not be trivial to administrate; but I think it's still feasible. Of course, local initiatives would still be voted only by local residents/taxpayers.)

 

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

William Waugh 

--

[Jameson Quinn]

I think Phragmen, Thiele, and Ebert are a clear theoretical advance, and I really like the work Toby has been doing thinking and writing about these. But I don't think that voters would accept them as a first step into proportionality; too hard to understand. 

[William Waugh]

 In the discussion on Canada, some have mentioned using methods that require approval ballots, but wanting to offer the expressiveness of score ballots, and solving the mismatch by mathematically dividing a vote into pieces that would all use approval-style indications.

Another way to convert would use a pseudorandom variable to convert the scores to approvals. It could be dicey to sell, because a listener could think that the result should depend on just the votes, and because it sounds like an opening to rigging the tally if the pseudorandom process is not fully exposed. But perhaps the concern could be addressed by publishing the algorithm, choosing a relatively simple one, and specifying the seed in a deterministic way e. g. tally the ballots within each riding in order of serial number and let the seed for each calculation come from the previous, so the entire tally can be reproduced. To counter the intuition that making the result depend on anything other than the votes is bad, we could reply that for a large electorate a study from a statistical viewpoint could show that much variation in the results coming from the pseudorandom calculations would be unlikely, and we could point out that every voting system can result in a tie, and that even though that is unlikely, it is still theoretically possible, and that in any number of polities the election laws call for a random process to resolve a tie, and so therefore the idea of including limited randomness is not new.

[William Waugh]

On Fri, Nov 20, 2015 at 6:25 PM, Jameson Quinn  wrote:

2015-11-20 18:18 GMT-05:00 William Waugh:

How about use separate sections of ballot for the local election and 
for the top-ups? Wouldn't that simplify the reasoning necessary to 
understand the properties and likely effects of any proposed system? 

--sure, but now your ballot is more complicated. It now has 2 parts 
instead of 1, and the 2nd part is MUCH larger.

Let me add on in this context, that what is easier to reason about, is also easier to explain, and what is easier to explain, is easier to sell.

I have been skimming the discussion and if I understand right, some of your proposals would limit the set of candidates a voter would have to consider for the top-up phase to those running in a "region" of five ridings or so. So that would reduce the number of candidates the voter would have to consider.

Another way to limit the physical size of the top-up section of the ballot would be to leave off listing the candidates, but permit the voter to enter a code number, or a few. This would work for voters willing to research ahead of time to decide which candidacies to support, or voters who trust the advice of party operatives or groups advocating on policy. In some variants, a code number could represent a slate of candidates with scores rather than a single candidate.
 

I've made proposals based on both of these possibilities, since at least when I created PAL over 2 years ago. My current thinking is that it's simpler (and probably an easier sell) to just let voters opt to get a ballot from another riding.

(In an American context, where you have multi-page ballots including local initiatives and school board/sheriff/judge/clerk/dogcatcher, this "select your ballot" would be by race, which would not be trivial to administrate; but I think it's still feasible. Of course, local initiatives would still be voted only by local residents/taxpayers.)

If the voter is to be offered a choice of ballots, some variants, possibly worth consideration, would offer them a broad choice only for the top-up portion; their only choice for the local part of the election could be the ballot fragment for the riding in which they live. I read the arguments for a complete choice, but nevertheless, I suggest a full decoupling between the local aspect and the at-large (or quasi-large) aspects. Being able to choose the whole ballot from another riding would couple the voter's expression of opinion on who should represent the local interests to her expression of opinion on who should represent national, regional, or provincial interests. I contend that such coupling makes the proposals harder to understand than necessary.

[William Waugh]

It may be worth paying attention to the fact that if a scheme offers the voter a choice of ballot portion from one riding or another, the voter is not only choosing from which set of candidates the voter can vote for, but also from which set of candidates the voter can vote against.

[Clay Shentrup]

On Friday, November 20, 2015 at 3:27:22 PM UTC-8, Jameson Quinn wrote:

I think Phragmen, Thiele, and Ebert are a clear theoretical advance, and I really like the work Toby has been doing thinking and writing about these. But I don't think that voters would accept them as a first step into proportionality; too hard to understand.

Thiele as in 0-1 RRV? Too hard to understand? Radically simpler than STV. 

[Toby Pereira]

On Friday, 20 November 2015 23:27:22 UTC, Jameson Quinn wrote:

I think Phragmen, Thiele, and Ebert are a clear theoretical advance, and I really like the work Toby has been doing thinking and writing about these. But I don't think that voters would accept them as a first step into proportionality; too hard to understand. 

Thanks for the praise. I do think there's a possibility that people wouldn't accept a score system as a first step into proportional representation, but not necessarily because it's too hard to understand. I don't think most people really understand exactly what's going on in STV either. I think the fact that they are largely untested is a bigger potential problem. And you said yourself about vulnerabilities to strategic voting. Personally I don't think it's likely to be that great a problem. The later-no-harm criterion is also failed for single-winner score, and that seems to behave generally very well. Plus my general feeling is that PR methods are likely to converge rather than diverge relative to single-winner methods - i.e. agree with each other more.

Plus my latest proposal above (which I think is pretty good) that uses Warren's 13-5 thing works much better with score ballots than ranked ballots, even if this is then converted into a more STV-like method.

[Toby Pereira]

On Friday, 20 November 2015 23:35:19 UTC, William Waugh wrote:

 In the discussion on Canada, some have mentioned using methods that require approval ballots, but wanting to offer the expressiveness of score ballots, and solving the mismatch by mathematically dividing a vote into pieces that would all use approval-style indications.

Another way to convert would use a pseudorandom variable to convert the scores to approvals. It could be dicey to sell, because a listener could think that the result should depend on just the votes, and because it sounds like an opening to rigging the tally if the pseudorandom process is not fully exposed. But perhaps the concern could be addressed by publishing the algorithm, choosing a relatively simple one, and specifying the seed in a deterministic way e. g. tally the ballots within each riding in order of serial number and let the seed for each calculation come from the previous, so the entire tally can be reproduced. To counter the intuition that making the result depend on anything other than the votes is bad, we could reply that for a large electorate a study from a statistical viewpoint could show that much variation in the results coming from the pseudorandom calculations would be unlikely, and we could point out that every voting system can result in a tie, and that even though that is unlikely, it is still theoretically possible, and that in any number of polities the election laws call for a random process to resolve a tie, and so therefore the idea of including limited randomness is not new.

If you're doing a random thing, you could run it more than once to demonstrate that is generally gives the same result (assuming it does). 

[Jameson Quinn]

I've come up with a name for my latest proposal in this thread, and written a pitch for it on Quora:

EMFATICS voting. This is a novel proposed proportional representation system, using the best aspects of existing systems, which ensures accountability and proportionality with simple ballots. The initials stand for:

EM: Extra Member. 2/3 of the seats are determined locally by votes from a given riding/district; the other 1/3 are "extra members" allocated regionally to ensure a proportional result. (This kind of system is often called "MMP", mixed member proportional).

FA: Free Adherence. By default, you vote in your own district. But if you don't like any of the candidates in your district, you are always free to cast your vote in some other district. The more voters choose to vote in a district, the more likely that district will get extra members.

TI: Transfer Indepently. Your vote automatically helps make sure your favorite party gets its fair share of seats. After all the parties all have their minimum fair share, each party will still have some leftover votes. Parties with the smallest leftovers will pass them to those with bigger leftovers. Each voters' preferred candidate is the one who decides how that vote is transferred; in other words, for transfers, the vote is delegated.

CS: Combined Score. You can give each candidate in your district (or whichever district you choose to vote in) a score of 0, 1 , or 2, corresponding to unapproved, approved/acceptable, or preferred. Within each district, the highest total score wins; then, within each party, the highest scorers are first in line for the "Extra Member" seats.

To illustrate this, take the example of the US state of Arizona. Its population is enough for 9 seats in the House of Representatives. Currently, that means it has 9 districts:

(You can see the effects of gerrymandering in the bizarrely-shaped peninsulas and isthmuses of these districts, especially the 4th and 9th.)
Under EMFATICS, it would have 6 districts and 3 "extra seats". Each party would get a seat (local or extra) for every full 10% of ballots which preferred it (that is, 100% divided by the number of seats plus one).
Say Republicans scored highest in 4 of the district races, and Democrats scored highest in 2. Meanwhile, the overall breakdown of preferred votes (scores of 2) was:
43% Republican (4 seats; already taken; remainder 3%)
36% Democratic (3 seats; 2 taken; remainder 6%)
11% Libertarian (1 seat; remainder 1%)
7% Independent (remainder 7%) (Note: this is a single candidate. If there were several independents, each would be considered as their own separate party).
3% Green (remaider 5%)

After the local winners are seated, the Democrats and the Libertarians each have enough votes for one extra seat; these go to the highest-scoring candidates available in each case.
Now the non-winning candidates from the party with the lowest remainder — the Libertarians — must choose how transfer their remainders to other parties. Each non-seated Libertarian candidate controls a number of votes proportional to how many voters preferred them; they decide which party to give to. For simplicity, let's say they all choose the Republicans, so the Republicans go from 3% remainder to 4%.
The next lowest remainder is now the Greens. Let's say one Green gives her 1% of votes to the Independent (bringing him to 8%), while the other Greens give their 2% to the Democrats (bringing them to 8% too).
The next lowest remainder is the Republicans. Let's say they all give their 4% to the Independent. That brings him to 12%, more than enough for the final seat. He is seated, and the process is complete. So, the final totals are: 4R, 3D, 1L, 1I. 

...

What are the advantages of this system?
Fully proportional: Almost all voters are guaranteed that their vote will help elect a representative that they supported, directly or indirectly. In the example above, only the 3% Greens went entirely unrepresented. 
Voters decide: There are no "safe seats" that are not accountable to the voters. Because a party's extra seats always go to whichever of its candidates have the highest scores, and because voters always have the option to vote in another district if there are no good candidates locally, corrupt party insiders have no way to guarantee reelection or shield themselves from accountability.
Simple ballots: You never have to look at more than one district worth of candidates. Any voter who votes one candidate as "preferred" can be assured that their vote will help get that candidate's party fair representation.
Independents get a fair chance: Even if an independent doesn't quite win their district, they still have a fair chance to get one of the "extra member" seats. But that choice is made by actual candidates from other parties choosing to give their remainder votes to the independent, not just by a mathematical formula. So a centrist independent with crossover appeal who got 45% preferences locally might be enough to get a seat; but an extremist with the same local total would not.

Note: This system is related to SODA, an earlier proposal of mine. I have been trying to simplify yet further, yet keep the main advantages, and this is the result.

.........

(no longer part of the Quora answer)

Note that the above does not deal with "overhang". Any party with more local seats than they deserve proportionally would still get those seats, but have zero remainder votes, and would not be eligible to get vote transfers. The quota would then be recalculated as the number of votes for all other parties, divided by the number of remaining seats plus one.

I believe that's the only gap in the above. In previous posts I've suggested "optional" rules to ensure that candidates are not elected without enough direct votes, and that strong local candidates cannot be shut out by non-local votes. I think both of those things are still reasonable ideas, but I've left them out of the above for simplicity. 

--

[Bruce Gilson]

In general, I like PR systems, and because I am a Republican who lives in a district that is strongly Democratic (the Dem has at times won with 77% of the vote, and even after the last redistricting, which made it less Democratic to force a Republican out of a seat from a neighboring district, my district still went 60% for the Dem in a 4-person race) so I feel I would never have a representative who reflects my own political thoughts, I tend to like systems such as SODA (and therefore am favorably inclined toward Jameson's EMFATICS). However, I wonder what is the chance that it would have the effect of leading to districts (ridings in the Canadian-oriented language of Jameson's proposal) with a representative who does not really represent his district. (What happens if such a large number of voters vote for someone outside their district that the winner of that district is elected, not only with a small minority of the voters in the district, but in a system where all the candidates in the district together receive less than a majority?)

Just a question.

[Jameson Quinn]

To respond to Bruce's question:

Say that it's Virginia 1998 (11 districts), and Ralph Nader is running in district 1, and he gets 10% of votes from around the state. From inside the district, say that there's 3% (of the state) for the Republican and 4% for the Democrat (plus 1 of the 10% for Nader). So the Democrat "should" win the district, but she's totally overwhelmed by the external votes for Nader.

Well, we know that the Democrats have at least 4% of statewide total, and Nader has at least 2% leftover to pass on. Chances are that the Democrats will end up with at least one of the 4 "extra member" seats. And chances are that any Democrat with over 4% will have won in their district (since we know that there's an average of about 7% of the state per other district), so that the 4% the district 1 Dem has will be enough to get her a seat.

Now, it's possible that the Dems filled up their proportional seats with local wins. It's even possible (though unlikely) that there's another Dem somewhere else who got more than 4% but didn't win locally. So the system does not guarantee that the district 1 voters get their local plurality candidate elected. But it is pretty likely. And if they don't, they will at least be proportionally represented by candidates from the same party from somewhere else.

I think that the system should require the parties give their winners "territories", as in SODA, so that each district is in the territory of one winner from each winning party. But that can be left up to the parties.

Does that answer your concern?

2015-11-23 19:16 GMT-05:00 Bruce Gilson <brg...@gmail.com>:

In general, I like PR systems, and because I am a Republican who lives in a district that is strongly Democratic (the Dem has at times won with 77% of the vote, and even after the last redistricting, which made it less Democratic to force a Republican out of a seat from a neighboring district, my district still went 60% for the Dem in a 4-person race) so I feel I would never have a representative who reflects my own political thoughts, I tend to like systems such as SODA (and therefore am favorably inclined toward Jameson's EMFATICS). However, I wonder what is the chance that it would have the effect of leading to districts (ridings in the Canadian-oriented language of Jameson's proposal) with a representative who does not really represent his district. (What happens if such a large number of voters vote for someone outside their district that the winner of that district is elected, not only with a small minority of the voters in the district, but in a system where all the candidates in the district together receive less than a majority?)

Just a question.

--

[Bruce Gilson]

On Mon, Nov 23, 2015 at 9:27 PM, Jameson Quinn <jameso...@gmail.com> wrote:

​[..

.

​]

I think that the system should require the parties give their winners "territories", as in SODA, so that each district is in the territory of one winner from each winning party. But that can be left up to the parties.

​I would prefer that such a territorial assignment be explicitly written into the rules. That way, each voter has a representative that he can consider "his."​

[Toby Pereira]

I'm going to clarify my latest method, based on Warren's 13-riding district system with 5 top-up MPs. The numbers 13 and 5 aren't essential parts of the system, but it wouldn't work with too many ridings as the ballot paper would get too large.

1. Each riding's ballot paper would list the candidates standing in that riding, along with any party affiliation, with a way to give the candidate a score. This could be a box to enter the score, or a list of potential scores in a line with the voter to circle one.

2. Next to or underneath each party candidate, there would be a separate mention of their party, which can be separately scored.

3. Underneath the list of the ridings candidates (possibly separated by a line) would be a list of all independent candidates standing in the whole region as well as a list of all parties fielding candidates elsewhere in the region but not in this particular riding. These would also be scorable by voters.

4. Voters can give scores to as many of the listed candidates or parties as they like, and can ignore as many as they like. The score for a candidate and their party do not have to be the same.  If a voter gives a score to a candidate but not their party or vice versa (so one is left blank as opposed to being given an explicit zero), then by default the same score is applied to both. Other than this, blanks are taken to be zeros.

5. The scores given to the ridings candidates on the ballots from that particular riding are added up and the highest scoring candidate is elected as MP for that riding.

6. The top-up phase commences. In addition to the scores given to the candidates from voters in their ridings, all scores are now considered. Any score given to a party counts for all candidates in that party from the region, apart from the scores explicitly given to ridings candidates if they are different. The scores explicitly given remain as they are. Scores given to independent candidates from outside their riding are also now considered.

7. The rest of the seats are now allocated using a proportional score system. The seats that have already been allocated are taken into account, so that the result is the most proportional it can be given those already elected.

END

Just a note on scoring candidates and their parties separately, I'm not sure what the cleanest look would be. Such as:

Candidate name BOX Party name BOX

or

Candidate name 0, 1, 2, 3, 4, 5, 6, 7, 8, 9

Party name         0, 1, 2, 3, 4, 5, 6, 7, 8, 9

or even

Candidate name 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 Party name  0, 1, 2, 3, 4, 5, 6, 7, 8, 9

It makes sense to try and get them on the same line if possible because the ballot would have quite a few names on it. I still think the size would be within acceptable limits without having to go to delegated votes/asset voting or only allowing voters to rate candidates in their own riding.

[Warren D Smith]

Toby, your systems sounds like my 13,5 system, except with cleanup of
ballots and more precise notions of how the ballots should look, and with
ability of voters to score candidates outside their riding.

I actually was thinking of pretty much exactly the same system as that
myself, but
what I do not like about it, is the large hairy ballots.

[Toby Pereira]

The method might require further modifications if the ballot is too large. But it seems we've ruled out asset/delegation. So I would suggest reducing the region size. We could have maybe 8 ridings and 4 top-up seats (so a 2 to 1 ratio). This might seem like it gives us a reduction in proportionality, but that isn't necessarily the case. If having 13 ridings requires a ballot with only the local candidates, then what might superficially look like a high level of proportionality would in fact not be (see my earlier postings on this). Allowing voters to vote for candidates/parties outside their own riding is required to make the proportionality work. We could go down to 8 and 4 as I have just suggested or maybe even push it to 6 and 3. I think that would probably be the limit though for a decent level of proportionality.

[Andy Jennings]

Jameson,

I do like this EMFATICS system.  Nice and simple.  Basically, MMP w/ Asset (whoever has the smallest total must transfer first).  Right?

Nice.

~ Andy

[Warren D Smith]

On 11/24/15, 'Toby Pereira' via The Center for Election Science

<electio...@googlegroups.com> wrote:
> The method might require further modifications if the ballot is too large.
> But it seems we've ruled out asset/delegation.

--we have? I thought JQ's EMFATICS had asset/delegation
in it? (It is, I assume, living off the phat of the land? :)
I'm just anti-complexity. Asset adds some extra complexity, so
do large ballots. But those may be worth the cost. It's all a
tradeoff, you have
to keep weighing some cost versus some benefit in a very hard to quantify
manner. Guess a little too wrong and the result is failure. I'd like these
tradeoffs to be assessed more quantitatively if we can.

> So I would suggest reducing
> the region size. We could have maybe 8 ridings and 4 top-up seats (so a 2
> to 1 ratio). This might seem like it gives us a reduction in
> proportionality, but that isn't necessarily the case. If having 13 ridings
> requires a ballot with only the local candidates, then what might
> superficially look like a high level of proportionality would in fact not
> be (see my earlier postings on this). Allowing voters to vote for
> candidates/parties outside their own riding is required to make the
> proportionality work. We could go down to 8 and 4 as I have just suggested
> or maybe even push it to 6 and 3. I think that would probably be the limit
> though for a decent level of proportionality.

--yah, well, this is an example of intuition, not quantitative.

[Jameson Quinn]

Yes, emfatics has delegation, but intraparty transfers are handled automatically; only interparty uses delegation.

I believe that any system has failure modes. The point is to make it so failure is the exception, not the rule as with SMP. The failure mode of 13/5 is still gerrymandered safe seats. Much less so than plurality, but still an issue, I think. And there's also the corresponding charybdis of a profusion of independent candidates who ruin propotionality.

The failure mode of EMFATICS is larouche-like party hijacking. There may need to be an approval threshold partisan primary.

Still, I think both of these have been pretty well-optimized.

[Warren D Smith]

> The failure mode of EMFATICS is larouche-like party hijacking. There may
> need to be an approval threshold partisan primary.

--what is "larouche-like party hijacking"?

[Jameson Quinn]

Lyndon LaRouche is a conspiracy theorist with a cult following who sometimes runs as a supposed Democratic candidate for various offices. He is despised by most actual Democrats (as well as most actual Republicans and Independents), but sometimes he gets some votes from people who don't know he's not a Democrat.

[Warren D Smith]

On 11/25/15, Jameson Quinn <jameso...@gmail.com> wrote:
> Lyndon LaRouche is a conspiracy theorist with a cult following who
> sometimes runs as a supposed Democratic candidate for various offices. He
> is despised by most actual Democrats (as well as most actual Republicans
> and Independents), but sometimes he gets some votes from people who don't
> know he's not a Democrat.

--well, how does he acquire the Dem Party nomination for those offices?

[Toby Pereira]

On Wednesday, 25 November 2015 01:56:48 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 11/24/15, 'Toby Pereira' via The Center for Election Science
<electio...@googlegroups.com> wrote:
> The method might require further modifications if the ballot is too large.
> But it seems we've ruled out asset/delegation.

--we have?  I thought JQ's EMFATICS had asset/delegation
in it?   (It is, I assume, living off the phat of the land? :)
I'm just anti-complexity.  Asset adds some extra complexity, so
do large ballots.  But those may be worth the cost.  It's all a
tradeoff, you have
to keep weighing some cost versus some benefit in a very hard to quantify
manner.  Guess a little too wrong and the result is failure.   I'd like these
tradeoffs to be assessed more quantitatively if we can.

I just thought I remembered you saying somewhere that asset/delegation would probably be rejected for being too complicated. Obviously delegation featured in some of my ideas too but I thought I'd move away from that in light of this comment. But yeah, I understand that it's all about the best trade-off.

 

>  So I would suggest reducing
> the region size. We could have maybe 8 ridings and 4 top-up seats (so a 2
> to 1 ratio). This might seem like it gives us a reduction in
> proportionality, but that isn't necessarily the case. If having 13 ridings
> requires a ballot with only the local candidates, then what might
> superficially look like a high level of proportionality would in fact not
> be (see my earlier postings on this). Allowing voters to vote for
> candidates/parties outside their own riding is required to make the
> proportionality work. We could go down to 8 and 4 as I have just suggested
> or maybe even push it to 6 and 3. I think that would probably be the limit
> though for a decent level of proportionality.

--yah, well, this is an example of intuition, not quantitative.

Right. I suppose the numbers I picked don't have proper mathematical/experimental backing behind them. But I think it would make sense to use the largest number of ridings possible before the ballot gets ridiculously big. You said yourself 13 is likely to be to big. But I'm not sure I can do much more than use intuition about what ballot size would be too big without a very big survey.

[Warren D Smith]

well, I based my choices of #topup MPs and such on data from here:
http://rangevoting.org/CanadaSeatsVotes.html

[Toby Pereira]

OK, but in your report, 33% of the seats would have needed to be top-up seats to ensure decent proportionality. You've suggested it might be less with score voting, but you also lose a certain amount of proportionality by switching from national to regional and having a smaller number of seats. So I would suggest keeping the 33% figure, which also would keep voters content that most representatives are local representatives. And if we want a ballot where voters can vote for candidates outside their riding (my proposal that you also came up with something very similar to) then we need to limit the number of ridings per ballot. As said, you were concerned that 13 would be too large in this case. 12 would be very similar, so the choices are probably 10 (with 5 top-ups), 8 (4 top-ups) and 6 (3 top-ups).

[Jameson Quinn]

I understand that you're replying to Warren's idea, but I have proposed a way for voters to vote outside their riding without making the ballot any more complicated: voters choose which ballot to use.

--

[Warren D Smith]

On 11/25/15, Jameson Quinn <jameso...@gmail.com> wrote:

> I understand that you're replying to Warren's idea, but I have proposed a
> way for voters to vote outside their riding without making the ballot any
> more complicated: voters choose which ballot to use.

--would that whole idea meet immediate massive opposition?
("Why're these Mexicans choosing MY representative?")

It is hard for me to judge the possible goofy political roads this
could be channeled into on the way to the gallows, but that one seems
plausible ... right now Trump is making major political progress due
(?) largely to xenophobia of a related but probably less justified
ilk.

[Toby Pereira]

I don't think being able to vote outside your riding is a terrible idea, but I don't think it solves all the problems either. As Warren says, people might complain about outsiders electing their representative. And maybe I'm getting a bit carried away, but I can imagine social media campaigns to oust particular politicians by getting the ballot for their riding, and then a counter-campaign - the result being that a ridiculous number of people vote in one riding.

Also, for the system I've described (i.e. with no asset/delegation), we're still left with the problem that the ridings are unconnected and the proportional top-up phase wouldn't work properly. I think it only makes sense for systems that has some way of transferring votes between candidates, which obviously emfatics does but what I have described doesn't.

[Jameson Quinn]

2015-11-25 18:58 GMT-05:00 'Toby Pereira' via The Center for Election Science <electio...@googlegroups.com>:

I don't think being able to vote outside your riding is a terrible idea, but I don't think it solves all the problems either. As Warren says, people might complain about outsiders electing their representative.

In practice, a representative that was strong enough locally would end up getting seated one way or the other; with either a "local" or "extra" seat. If external votes put some other candidate above them, they'd still be first in line for an extra seat.

...

OK. I've been thinking about rules to patch this. And they exist, but they end up being too complicated. So fine. Here's a change which makes a new canonical version of EMFATICS. Anything referring to my old proposal should call it "EMFATICS version 1". 

Local winners to be determined by highest score local votes only. If no local candidate gets a score equivalent to preferences from at least 1/3 of the voters from that riding (or approval from 2/3), there is no local winner for that riding, and the seat is allocated as "extra". 

Extra seats for a party are still given in order of highest total score, counting both local and nonlocal votes.

So instead of nonlocal winners winning locally and local winners being first in line for extra seats, it's the other way around. It makes a little bit less sense in a pure vote-theoretic perspective, but a lot more sense intuitively.

 

And maybe I'm getting a bit carried away, but I can imagine social media campaigns to oust particular politicians by getting the ballot for their riding, and then a counter-campaign - the result being that a ridiculous number of people vote in one riding.

This is not really a problem. The "ousting" part probably would not have worked even in version 1, because of the "first in line" thing; and it has no chance in version 2. And the countercampaign is useless in both versions; if two candidates in a riding both get a quota of direct votes, they are both guaranteed seats, so no "ousting" is possible.

[Warren D Smith]

In Asset voting voters must supply scores which sum to a fixed value, whereas in
EMFATICS a voter could deliver as many "2" scores as she wishes.
The asset "juice" is thus not starting out equally; if the Democrats run more
candidates they'll win more seats, i.e. EMFATICS
is not at all cloneproof. Doesn't this seem a devastating flaw?

[Warren D Smith]

Perhaps Jameson had intended (but forgot to say it) that each EMFATICS
voter award exactly one "2" ("preferred") score. That would repair the
cloning problem I pointed out last post.

[Jameson Quinn]

I believe I said that you delegate in equal fractions to all candidates you give a 2 to. Certainly I intended it.

[Warren D Smith]

JQ: I believe I said that you delegate in equal fractions to all candidates you

give a 2 to. Certainly I intended it.

--WDS: not in the text I've got from you! But anyhow, fine.

JQ re EMFATICS 1:

In practice, a representative that was strong enough locally would end up
getting seated one way or the other; with either a "local" or "extra" seat.

WDS:
I must dispute this claim.
See, it often happens that a party, most often the leading party, is
overrepresented in local seats. It then will get (and deserve) ZERO topup seats
(or near zero).

So suppose Evil Rival Party wants to get rid of the prime minster
(from the leading party).
They organize campaign of voters Canada-wide to vote external to their
ridings, in the PM's riding, for his rival. This also might well
happen naturally with no need for organized campaign. Many voters
will feel their riding is "safe" hence they might as well vote
strategically thus.

Result: PM loses his local set, does not regain it via topup since his
party has no topup seats.

This is reminiscent of what I was calling "targeted killing" about
Balinski fair majority: voting. It is not as severe for EMFATICS 1 as
for Balinski (since Balinski caused massive amplification and there
was no defense possible) but it is still plenty bad.

JQ then came up with EMFATICS 2:

Local winners to be determined by highest score local votes only.

--WDS: good, that removes the targeted killing problem.

JQ: If no local candidate gets a score equivalent to preferences from

at least 1/3 of
the voters from that riding (or approval from 2/3), there is no local
winner for that riding, and the seat is allocated as "extra".
Extra seats for a party are still given in order of highest total score,
counting both local and nonlocal votes.

--WDS: doesn't that mean many ridings will end up with no MP (no
local, also no topup)?
That could be unpleasant. Yes, they did not like any candidate...
but still could be unpleasant to be unrepresented. It happens quite
often in approval voting election poll studies that nobody gets
majority-approved.

To avoid that, perhaps it would be better to add an explicit "none of
the above" choice; or
to lower the "1/3" to 1/10, or 0, or something.

[Warren D Smith]

Jameson, I made an EMFATICS page here
http://rangevoting.org/EMFATICS.html
please let me know if you approve it, want changes, etc etc.

I currently intend to add this as an additional recommendation in
CanadaOverview.

Also I intend to add Toby Pereira's suggested system ditto.

However, I want to have another think about CanadaOverview
in light of everybody's ideas, first...

[Jameson Quinn]

Warren: you've made several edits to EMFATICS. Some of them are just language changes, and I think the majority of these are improvements, and the ones that in my opinion aren't, aren't too bad.

The one substantial change you've made is to add NOTA. I think this is a good idea, and am willing to add that to the "canonical" definition of EMFATICS. However, if you do use NOTA, I believe that voting in another riding should automatically add 2 to NOTA's score in your native riding.

Here's your "criticisms" section, with my responses:

Looking up numeric codes for out-of-district voting (or acquiring ballots in other ridings): This may be too difficult for voters, and the fact many races are generally printed on the same ballot for a riding (not just the federal MP race) makes it difficult to use the "just use the foreign riding's ballot" solution. Also many people vote by mail (including in Canada), complicating matters further.

I believe that this would be a solvable logistical problem. I'm going to give a sketch of a solution here. I'm sure my sketch could be improved upon.

Consider a by-mail voter in an election that consists of mutiple sub-elections (for instance, for national, regional, and municipal representatives, as well as some single-winner races and/or referenda, in a place where all three levels of representation use EMFATICS). This voter would get a ballot of at least 4 "pages": one for each of the EMFATICS races, and one for the other races. I put "pages" in quotes because there could be two EMFATICS races on opposite sides of a single sheet of (sturdy) paper. 

It would be desirable to keep EMFATICS races from running to more than one page long; for that purpose, you could have a fixed number of "preprinted" parties and independents, with write-in slots for candidates lower down. I estimate that, on a "legal sized" paper, with a reasonably large font, spacing, and header, you could fit about 25 candidates per side. That could be:

2 slots each for the 6 parties which got the most votes in the prior election

4 slots for "new parties" which got the most regionwide petition signatures.

9 slots for the independents who got the most local petition signatures.

Each ballot line would be formatted as:

Party name / symbol... Candidate Name/info .......check boxes

The ballot would be folded over and so as to make a "pocket" on the left side. The "pocket" would be sealed at top and bottom by tabs which folded around. There would be party names written on the outside of the pocket. There would be further paper extending out from the pocket which was not attached at top and bottom, which had the local candidate names. 

Voters who wanted to vote in other ridings would remove the part of the paper with local candidate names. This would leave a ballot with lines looking like this:

Party name / symbol... Blank line................check boxes

Voters could then obtain official "supplements" for the riding of their choice, which would fit into the pocket and align with the check boxes. Providing falsified supplements would be a crime. But even if you did do it, voters would still be able to see the party designations printed on the ballot.

At the top of the ballot, voters would hand-mark the code for their riding. This code would be visible on the removable paper or the supplements, so it would just be a matter of copying a 2-digit riding number into boxes a few centimeters from where you see it written. Voters who did not explicitly mark a code for their riding would have their partisan preferences count, but their individual preferences would not count.

There would also be official apps for major smartphone platforms which could look at a ballot using the phone's camera and say whether it was valid and who it would vote for. These apps would notice and give a warning if a non-official supplement insert was used. There would be ways to get access to devices with these apps; for instance, at libraries, post offices, etc. In such locations, there would be a way to scan your ballot without having to hold the device by hand or worry about shadows.

...I could go on and add more details, but you get the idea. I'm also sure that the above could and would be improved.

 

Independents would presumably be unfairly disadvantaged in this system, as it might be unrealistic to list all of them as well as the parties, and the explicit use of parties in this system causes a built-in bias against independents, as well as offending some ("as a matter of principle, voting systems should not ever need to refer to parties"). In its defense, though, EMFATICS does try keep its party-usage low and still ought to produce decent results even if there were no parties and all candidates ran independent.

I think EMFATICS is entirely fair to independents. I also think that, though such fairness is important, it is impossible to completely satisfy everyone; I am not concerned with those who'd stand on empty principles and ignore real attempts to address their concerns.

 

With only "two and a half" score levels (0,1,2 with at most one use of 2 per voter) EMFATICS inherently loses accuracy versus systems employingscore voting with, say, 0,1,2,3,4,5,6,7,8,9 as allowed scores. It is known that some voters in real elections genuinely needed at least 7 score levels to express their opinions, and that approval voting has substantially worse Bayesian regret than score voting.

The benchmark for EMFATICS is not score, but asset. Too many fine distinctions would allow too much strategizing to elect pseudo-independents locally and then get more than a fair share at the wider level.

Also, I find that most people are satisfied with 3 rating levels. It doesn't exhaust the information they could express, but it allows them to express the most important things: whom they support the most, and whom they consider acceptable/unacceptable otherwise.

 

It is possible that NES strategy by voters about the 2's might yield enough artificial advantage for the top 2 parties to ultimately engender 2-party domination.

No way. The same two parties would not be the frontrunners in all ridings. Small parties could explicitly campaign to concentrate their regional votes on a single local candidate. Both of these could still happen even in an environment which was otherwise NES-dominated.

[Warren D Smith]

OK,
(1) I added Jameson's desire for "automatic NOTA=2 if
vote in foreign riding" to my EMFATICS page.

(2) I did nothing about Jameson's immensely long "logistics are easy"
discussion, but yes, it ought to be a solvable problem :)

(3) I want now to express possible disagreement with Jameson's cavalier "no way"
to my worry EMFATICS could engender 2-party domination via NES strategy.

See, in countries like Canada, the whole ball of wax, or anyway
massively important, is which party has control of parliament. E.g.
the Tories vs the Liberals.
A voter might consider that the most important thing. Therefore, the
voter gives her "2" to
the Tories, because that is the most important thing -- making sure
the Tories retain control of parliament. Say she does this
dishonestly, because really her favorite local candidate, and favorite
party, both are not Tory.

OK, now if enough voters think that way, then the Tories and Liberals
get practically all the "2" votes. (I consider this an entirely
plausible scenario. Not a cavalier "no way.")
As a result, we get 2-party domination in extra member seats. As a
result of that (over historical time) there will also be 2PD pressure
even in local MP seats... EMFATICS has some plurality-like character
causing 3rd parties to be disadvantaged in local MP races.

So, I think this all is a genuine worry. It is hard for me to assess
how serious a worry it is, and it might take 100 years to happen...
but I cannot just dismiss it. This same problem,
by the way, also arises in my own "asset used for top up" schemes.

Anyhow, now I've been thru this exercise of writing up EMFATICS, I
like it less than when I started. It's sort of been designed by a
"cuteness engineer." It looks very cute on
the surface, but then I came up with enough criticisms of it to start
getting unhappy. It's
like a veneer of cute covering underlying rot in the form of a lot of
property-sacrifice.

In the absence of convincing computer sims and/or convincing
historical evidence to quantify a multiplicity of what hopefully are
"minor" property sacrifices, I feel a lot happier if system has
provable properties, period. Yet another guy just came to me saying
he might write a computer simulator (so far, nobody who's said that
has actually done it and finished the job) for BR for multiwinner
elections, so that'll help if so...

[Neal McBurnett]

Thanks for putting up a page on EMFATICS, Warren (assuming Jameson approves of it)!

I like a lot of EMFATICS, but I think you're both downplaying the difficulty and costs involved in designing ballots which are easy to use. Current ballots are already often over-complicated. Many elections have had the outcome not reflect actual voter intent because voters were too confused to properly cast their ballots. Adding the space and explanatory requirements of score voting even with just 3 choices is a pretty big deal.

So Re: "All the criticisms in the preceding 3 paragraphs could be lessened in severity if EMFATICS were redesigned as follows: the allowed scores are now {0,1,2,...,M-1,M} where at most a single "M" (max) score is allowed per voter (M plays the role that "preferred" played in the original design). Then make M fairly large, such as M=10."

I think you should also discuss the impact on ballot design, voter confusion, etc.

Cheers,

Neal McBurnett http://neal.mcburnett.org/

[Warren D Smith]

what do Canadians want? Bad survey tries to answer:

http://rangevoting.org/WhatCanadaWants.html
http://rangevoting.org/AbacusCanadaERsurvey2015.pdf

[Clay Shentrup]

"all of them worded in an incredibly asinine manner."

Why do you sabotage your chances of being taken seriously? This kind of language is an especially huge liability for you since your site is just text, and doesn't provide any evidence of your being a "legitimate" ("vetted", whatever) organization.

[Warren D Smith]

On 12/2/15, Clay Shentrup <cl...@electology.org> wrote:
> "all of them worded in an incredibly asinine manner."
>
> Why do you sabotage your chances of being taken seriously?

--the wording in question is the question on p.8 of
http://rangevoting.org/AbacusCanadaERsurvey2015.pdf

Sorry, I have to wonder how anybody could produce a question wording
that completely asinine :) It's sort of a parody, it belongs on a
comedy show.
Those guys are professional pollsters or social scientists??

Anyhow, there are some good lessons to be learned from that poll in spite of
their efforts to self-sabotage it.

[Clay Shentrup]

On Wednesday, December 2, 2015 at 8:54:16 PM UTC-8, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 12/2/15, Clay Shentrup <cl...@electology.org> wrote:
> "all of them worded in an incredibly asinine manner."
> Why do you sabotage your chances of being taken seriously?

Sorry, I have to wonder how anybody could produce a question wording that completely asinine  :)

I understand that, and it's fine to vent about it here. But if you want your analyses to be taken as seriously as possible by whatever audience may come across it, it immensely benefits you to sound like a sane and reasonable person. When you use insults like "asinine", you come across as just some crazy person ranting on the internet.

My wife is a far more diplomatic person than me. She could probably run for office and do well. She has a master's in environmental management, and works in renewable energy procurement and policy for our local utility company. I read her these kinds of things and she just shakes her head and comments on how crazy you sound.

But the thing is, I know you're brilliant and that your ideas could completely change the world. But most people aren't like me and the other people on this list. Most elected officials already want to ignore your egghead math theories and focus on fundraising and moving their careers ahead. So you already have your work cut out for you. So then why be your own worst enemy? Just be a politician. Even if you think the authors are the world's greatest idiots, just use a polite and blunted criticism that they'll be more likely to respond to.

You are playing a game. The game has rules. There are these replication machines (humans) whose behaviors have been shaped by millions of years of evolution. You want to affect their actions. You may not like the rules of the game, but you have to play by them if you want to win. Do you want to win, or just get jollies out of castigating people who aren't as smart as you?

[Warren D Smith]

>> On 12/2/15, Clay Shentrup <cl...@electology.org <javascript:>> wrote:
>> > "all of them worded in an incredibly asinine manner."
>> > Why do you sabotage your chances of being taken seriously?
>>
>> Sorry, I have to wonder how anybody could produce a question wording that
>>
>> completely asinine :)

--Clay: how would you reword this?

[Clay Shentrup]

On Thursday, December 3, 2015 at 12:19:06 PM UTC-8, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

--Clay: how would you reword this?

I assume you're talking about the part that begins: "If someone asked you..."

It's kind of like having a question that reads: If someone were to ask you what your favorite color was, what would your answer be?

Yes, it is asinine. But it's probably not even worth your time to address that particularly idiotic question.

In general, you don't want to be so strident. For instance:

"This poll was badly written – and it also was conducted online, which is a bad way to conduct polls, although they claim their sample is representative anyhow."

Could be phrased more like:

"This poll contained several ambiguous and/or redundant questions. It was also conducted online, which raises concerns about its representativeness."

When you just assert that something is "bad", it sounds like you're not being objective and impartial.

"for four possible caring levels from most to least, all of them worded in an incredibly asinine manner."

Anyone reading this is likely to immediately stop and conclude you are a crazy person. State specifically what's wrong with how they were worded. Was it bad to use subjective terms instead of a simple linear numerical scale? Would the wording have been better if it was "not at all", "somewhat", etc? Just tell the reader what the problem is, as you see it. Sound calm and rational.

[Warren D Smith]

>"for four possible caring levels from most to least, all of them worded in an incredibly asinine manner."
>Anyone reading this is likely to immediately stop and conclude you are a crazy person.

--Fascinating. So you immediately conclude anybody using the word
"asinine" is crazy.

What is interesting about that behavior by you is, I conclude that
*you* are crazy.

But nevertheless, I invite you to suggest rewordings of whatever you
want, and if you succeed in finding a wording I like better, I'll
change it.
You haven't managed to do so, but you could.

[Clay Shentrup]

On Friday, December 4, 2015 at 6:16:40 AM UTC-8, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

--Fascinating.  So you immediately conclude anybody using the word "asinine" is crazy.

It depends on the context of course. Suppose you tell you're friend, "My wife's a bitch." He probably won't think you're crazy.

But suppose you say that to someone in a job interview. They'll think you're crazy.

Warren, this is a job interview. You want to think about how you'll be perceived if you're lucky enough to get your analysis read by a policy-maker or influential academic or whatever. Your goal is to sound sane, reasonable, and trustworthy. You DO NOT want to say things like "suicidal idiot".

[Toby Pereira]

Just to add this this where I say "The rest of the seats are now allocated using a proportional score system". Things seem to have got rather complex on that front (see multiple threads), but from where I'm sitting at the moment, I think the best compromise is probably to use Ebert or Phragmen used sequentially but with a within-party monotonicity add-on. That is to say that if the next seat would go to a party candidate who has a lower total/average score than someone else from their party, then the highest scoring currently-unelected candidate from that party would be elected next. I think monotonicity failures would be very rare anyway - you have to go out of your way to contrive cases, unlike with IRV. And I think it's a better compromise than using Thiele or some hideously complicated hybrid system. And electing sequentially should not sacrifice much proportionality.

On Tuesday, 24 November 2015 21:23:49 UTC, Toby Pereira wrote:

I'm going to clarify my latest method, based on Warren's 13-riding district system with 5 top-up MPs. The numbers 13 and 5 aren't essential parts of the system, but it wouldn't work with too many ridings as the ballot paper would get too large.

1. Each riding's ballot paper would list the candidates standing in that riding, along with any party affiliation, with a way to give the candidate a score. This could be a box to enter the score, or a list of potential scores in a line with the voter to circle one.

2. Next to or underneath each party candidate, there would be a separate mention of their party, which can be separately scored.

3. Underneath the list of the ridings candidates (possibly separated by a line) would be a list of all independent candidates standing in the whole region as well as a list of all parties fielding candidates elsewhere in the region but not in this particular riding. These would also be scorable by voters.

4. Voters can give scores to as many of the listed candidates or parties as they like, and can ignore as many as they like. The score for a candidate and their party do not have to be the same.  If a voter gives a score to a candidate but not their party or vice versa (so one is left blank as opposed to being given an explicit zero), then by default the same score is applied to both. Other than this, blanks are taken to be zeros.

5. The scores given to the ridings candidates on the ballots from that particular riding are added up and the highest scoring candidate is elected as MP for that riding.

6. The top-up phase commences. In addition to the scores given to the candidates from voters in their ridings, all scores are now considered. Any score given to a party counts for all candidates in that party from the region, apart from the scores explicitly given to ridings candidates if they are different. The scores explicitly given remain as they are. Scores given to independent candidates from outside their riding are also now considered.

7. The rest of the seats are now allocated using a proportional score system. The seats that have already been allocated are taken into account, so that the result is the most proportional it can be given those already elected.

END

Just a note on scoring candidates and their parties separately, I'm not sure what the cleanest look would be. Such as:

Candidate name BOX Party name BOX

or

Candidate name 0, 1, 2, 3, 4, 5, 6, 7, 8, 9

Party name         0, 1, 2, 3, 4, 5, 6, 7, 8, 9

or even

Candidate name 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 Party name  0, 1, 2, 3, 4, 5, 6, 7, 8, 9

It makes sense to try and get them on the same line if possible because the ballot would have quite a few names on it. I still think the size would be within acceptable limits without having to go to delegated votes/asset voting or only allowing voters to rate candidates in their own riding.

[William Waugh]

Is the collection of voting systems to suggest for Canada now settled?

[Warren D Smith]

On 12/14/15, William Waugh <2knuw...@snkmail.com> wrote:
> Is the collection of voting systems to suggest for Canada now settled?

--no -- not yet settled.

My MP accomplice and I have settled on this much: we think score-style ballots
probably should be part of the solution.

The question then is, if a PR system is to be used, then what should it be?
So far, no PR proposal using score-style ballots has managed to come along which
has left everybody completely satisfied and unworried.

----------

Re what I have been calling the "NESD property"
http://rangevoting.org/NESD.html
I recently ran into the following interesting study in Sweden:

Leif Lewin:
Majoritarian and Consensus Democracy: the Swedish Experience,
Scandinavian Political Studies 21,3 (Aug 1998) 195-206
http://rangevoting.org/Lewin1998ScandinavianPoliticalStudies.pdf

QUOTE:
"Minority governments have been by far the most common.
The average support enjoyed [by the govt] between 1920 and 1994 has been 41.5%
[and graphic shows it varied all over between about 10 and 90]."

Lewin also reports about a score voting study on p.199. (Warning:
Lewin uses incorrect terminology a lot!)
3588 random Swedish voters in year-1991 electoral surveys were asked
to rate the 8 parties on a -5 to +5 scale; 76% responded (both phone &
in person interviews used).
Lewin says it is a well known defect of unrated unranked voting systems
that there is no way to express intensity of preference. He is now going
to overcome that obstacle. He finds on comparing the PR official election
versus his ratings that -- if average party ratings were translated
into seat shares -- the ratings say that the two largest parties
("Social Democrats" & "Moderates") deserve to be weaker. The scorers
tended to exaggerate by scoring these two MAX and MIN or vice versa.
They then tended to score all remaining parties intermediately.
He claims this behavior is the main underlying reason for the discrepancy.
There is a graphic on p.200 summarizing study.

My point: this experimental finding suggests that voters will employ
"NES" (naive exaggeration) strategy, if voting using score-style ballots in a PR
system. That is, they artificially rate the two leading parties "MAX" and "MIN"
in some order, before scoring the others.

This NES behavior has also been heavily seen in Australia using
rank-order ballots.
Now

NASTY THEOREM:
If all voters behave in that NES way, then with IRV, or with STV-PR,
or plurality voting, or with many asset-voting-type
delegated-plurality-vote PR schemes, or with party-list, that will
FORCE two-party domination (2PD) down our throats.

Mind you, although some voters employ NES voting strategy, the
percentage of them
that do, varies considerably. In USA, Australia, and Malta it is very
large, causing
USA to have 2PD in plurality seats,
Australia to have 2PD in IRV seats but not in PR-STV seats, Malta to
have 2PD in PR-STV seats, but Ireland avoided 2PD in PR-STV seats, and
most party-list PR countries have avoided 2PD. But this all suggests
that even in the countries listed that avoided 2PD,
they still suffer more 2P than they really should.

So I think it is important to try to make the voting system obey the
NESD property, i.e.
that heavy NES voter behavior will *not* force 2PD.

If you demand NESD, it winnows out a large number of the PR proposals
we've been talking about.

[William Waugh]

On Monday, December 14, 2015 at 1:58:13 PM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

 

...

 

So I think it is important to try to make the voting system obey the
NESD property, i.e.
that heavy NES voter behavior will *not* force 2PD.

If you demand NESD, it winnows out a large number of the PR proposals
we've been talking about.

 

Thanks for the update. Interesting.

Is failure of NESD one of the reasons that RRV is not on the list?

Among the systems that conform to NESD, do any of them use separate sections of the ballot for the local election and the at-large top-up?

[Toby Pereira]

On Monday, 14 December 2015 18:58:13 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 12/14/15, William Waugh <2knuw...@snkmail.com> wrote:
> Is the collection of voting systems to suggest for Canada now settled?

--no -- not yet settled.

My MP accomplice and I have settled on this much: we think score-style ballots
probably should be part of the solution.

The question then is, if a PR system is to be used, then what should it be?
So far, no PR proposal using score-style ballots has managed to come along which
has left everybody completely satisfied and unworried.

It may be that no system will be good enough to do that, but I now have a fair amount of confidence in what I just posted in the other thread:

1. Elect sequentially using Ebert/Phragmen, but maintain monotonicity through approval removal when electing each candidate and restoring these approvals for subsequent candidate elections (to keep up overall proportionality).

or

2. If the above is intractable, Elect sequentially using Ebert/Phragmen but (if we are using the ridings/region system outlined in the Canada thread) have a party monotonicity add-on, so that if a party candidate is to elected, they will always be the most popular currently-unelected candidate from that party.

[Warren D Smith]

On 12/14/15, William Waugh <2knuw...@snkmail.com> wrote:

> On Monday, December 14, 2015 at 1:58:13 PM UTC-5, Warren D. Smith (CRV
> cofounder, http://RangeVoting.org) wrote:
>
> ...
>
>
>> So I think it is important to try to make the voting system obey the
>> NESD property, i.e.
>> that heavy NES voter behavior will *not* force 2PD.
>>
>> If you demand NESD, it winnows out a large number of the PR proposals
>> we've been talking about.
>>
>
> Thanks for the update. Interesting.
>
> Is failure of NESD one of the reasons that RRV is not on the list?

--I left RRV out because I felt it wasn't good enough.
There are optimizing systems which presumably do
at least as well as RRV (since RRV doesn't optimize) -- at the cost
of more computing, but we can afford it. And they seem simpler
to define, and less ad hoc, too. And RRV is ok on NESD.

> Among the systems that conform to NESD, do any of them use separate
> sections of the ballot for the local election and the at-large top-up?

--that complicates matters (in particular, less clear what NESD even
should *mean* in such a 2-part setting...)

[William Waugh]

On Monday, December 14, 2015 at 8:02:35 PM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

...

> Among the systems that conform to NESD, do any of them use separate
> sections of the ballot for the local election and the at-large top-up?

--that complicates matters (in particular, less clear what NESD even
should *mean* in such a 2-part setting...)

That the system doesn't bring about 2PD in the face of heavy NES in both parts? If the local part is Range anyway, exaggeration is not naïeve, but rather, rational and correct, so the local part wouldn't seem to contribute to the hazard. So, the system is NESD if it refrains from favoring 2PD in the face of heavy NES in the top-up part?

[Toby Pereira]

On Monday, 14 December 2015 18:58:13 UTC, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 12/14/15, William Waugh <2knuw...@snkmail.com> wrote:
> Is the collection of voting systems to suggest for Canada now settled?

--no -- not yet settled.

My MP accomplice and I have settled on this much: we think score-style ballots
probably should be part of the solution.

Is there a timeframe for this? When would you need to have a final proposal? Also, how likely do you think it is that you'll be able to influence the voting system? If it happened, Canada adopting a system of PR based on score voting could possibly be the best thing to ever happen in the world of countries' voting systems, but presumably it would have to pass over a lot of hurdles first.