Instant Score-off

341 views
Skip to first unread message

clay shentrup

unread,
Oct 12, 2014, 2:32:38 AM10/12/14
to electio...@googlegroups.com
Someone has recently proposed to me the idea (which has been discussed many times in the past) of doing a Score Voting plus top-two. Only instant. The ratings can be used for a head-to-head matchup.

This is actually more interesting than it at first appeared to me. While it is more complex than Score Voting, it enjoys the property of being much simpler than IRV (e.g. it is additive, using a comparison matrix), as well as being defensible against the common screed about being anti-majoritarian. You always get a "majority winner", even if that simply means a guarantee against electing the Condorcet loser.

There is also some sense in which it motivates a certain level of sincerity, given that your effect on the head-to-head matchup component is comparable in significance to your effect on picking the top two.

It does have the downside of not being computable on ordinary dumb totaling voting machines, but it could be viewed as a "pragmatic compromise" between the utilitarian crowd and the majoritarian crowd (however irrational the latter camp may be).

Thoughts?

Jameson Quinn

unread,
Oct 12, 2014, 9:07:57 AM10/12/14
to electionsciencefoundation
This is a pretty good idea. Outcomes would be good — possibly even better than score voting, in some strategy models — and it's easy to explain.

One problem I see is the issue of score granularity. If few scores were available — say, 0-5 — then you'd have to give up a significant amount of first-round power in order to get second-round expressivity. On the other hand, I think that many scores — such as 0-99 — can turn off voters by seeming too fidgety and arbitrary. 

For me personally, I'd be happy with a compromise where the available scores were (say) 0,1,25,50,75,98,99,100. But I'm pretty sure that idea would be a non-starter with your average voter ("Huh?"). Perhaps if you phrased it in terms of letter grades - ABCDF are worth 50,40,30,20,0, then '+' or '-' are worth 1 point up or down. 

(Usually, when I talk about letter grades, it's in the context of Bucklin systems, as a way of preventing voters from assuming that the system is Score. That's not the case here; this is a different use case.)

Jameson

--
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.

clay shentrup

unread,
Oct 12, 2014, 2:10:35 PM10/12/14
to electio...@googlegroups.com

On Sunday, October 12, 2014 6:07:57 AM UTC-7, Jameson Quinn wrote:
One problem I see is the issue of score granularity. If few scores were available — say, 0-5 — then you'd have to give up a significant amount of first-round power in order to get second-round expressivity.

I think most people would express this the opposite way. "If I give the top two the same score, then I have no effect on the second round."

A good counter to that is, "Sure, but if you like them almost equally, is it really that big a deal?"

A 0-9 range seems like a good compromise, although I still think a 0-5 scale is sufficient.
 
For me personally, I'd be happy with a compromise where the available scores were (say) 0,1,25,50,75,98,99,100. But I'm pretty sure that idea would be a non-starter with your average voter ("Huh?"). Perhaps if you phrased it in terms of letter grades - ABCDF are worth 50,40,30,20,0, then '+' or '-' are worth 1 point up or down.

Oh gawd.

Anyway, I'm pretty excited about this system. It's simpler than IRV (max of two rounds) and more expressive. And it resists arguments, by naive people who don't understand game theory (or "logic"), about bullet voting and other such bogeymen.

Toby Pereira

unread,
Oct 12, 2014, 3:06:53 PM10/12/14
to electio...@googlegroups.com
I think this is interesting, but I think you'd have to have something like a 0-100 range. If you're explicitly having the added complication of these two rounds, it would seem pretty weird to then have a scoring range that actively prevents people from participating properly in the second round. 0-10 doesn't seem enough for a system like this, and 0-5 I think would be awful.

Warren D Smith

unread,
Oct 12, 2014, 3:34:34 PM10/12/14
to electio...@googlegroups.com
Falls into all the usual paradoxes that runoff schemes usually get zapped by.
(Nonmonotonicity, fails to elect Condorcet winner, failure of participation,
failure of district partitionability.)
More complicated than plain range voting and probably performs worse, too.
A non-instant-runoff 2nd round would at least encourage voter honesty.




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

clay shentrup

unread,
Oct 12, 2014, 4:03:57 PM10/12/14
to electio...@googlegroups.com
On Sunday, October 12, 2014 12:34:34 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
Falls into all the usual paradoxes that runoff schemes usually get zapped by.
(Nonmonotonicity, fails to elect Condorcet winner, failure of participation, failure of district partitionability.)

Although your BR figures seem to speak well of it.
 
More complicated than plain range voting and probably performs worse, too.

But simpler than IRV and better too. Which is the relevant consideration, since there's no evidence that Score Voting will ever achieve political viability, due to the fact that social choice theory is too esoteric for non-geniuses to understand. E.g. they will always demand insane things like "majority winners".

A non-instant-runoff 2nd round would at least encourage voter honesty.

But would "cost more money" and require voters to vote twice.

And one could argue that this scheme is actually highly resistant to strategy, since you want to have a good chance of discriminating between the top two.

But the main point is that you should compare this to IRV more than to Score Voting.

Warren D Smith

unread,
Oct 12, 2014, 4:50:53 PM10/12/14
to electio...@googlegroups.com
> Although your BR figures seem to speak well of it.
> http://ScoreVoting.net/StratHonMix.html

--aha, but that is for a separate runoff round, not combined "instantly."
Which makes no difference for honest voters, but likely quite a bit for
strategic ones.

clay shentrup

unread,
Oct 12, 2014, 6:26:06 PM10/12/14
to electio...@googlegroups.com
On Sunday, October 12, 2014 1:50:53 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
--aha, but that is for a separate runoff round, not combined "instantly."
Which makes no difference for honest voters, but likely quite a bit for strategic ones.

Could you help us out with some R-IRV figures then?

Imagine the following reform campaign, aimed at people whose sensibilities would otherwise lead them to support IRV:

R-IRV has the following advantages:

1) Makes it easy to rate all the candidates, rather than just 3
IRV is hard in this regard, specifically because you can't give two candidates the same ranking, and so you have to try to make a ballot which makes that difficult for voters to screw up, ergo our 3-column "RCV" ballot, which restricts voters to 3 choices. But giving the same score to two candidates is fine with Score Voting, so you can use a simple ballot.

2) Simpler two-round tabulation
Think of the insanity you get in multi-candidate IRV races, like one that went on for 20 rounds of elimination in SF (http://www.sfelections.org/results/20101102/data/d10.html). R-IRV is a much simpler two-round process, more like a true runoff: there are just two second-round contestants, and we elect the majority-preferred between them.

3) Results are easier to understand at a glance

Would you rather see a view like this for each round of results...

or this holistic view of Score Voting results?

Especially if you're a minor party or independent voter!

4) Results are precinct-summable
Precincts can each report a simple total of points for each candidate, as well as head-to-head matrix of ordered preferences. There's no need to send the full ballots to a central tallying location.

5) Guarantees a majority winner in just one round of voting! Magic!

-----------

Now, Warren, would you admit that this might have some political advantages in the world of elected officials and activists who don't know Bayesian Regret from their left foot? Is it something you could support as a big improvement over the status quo? Could you help us with some Bayesian Regret figures for this system?

Thank you!

Mark Frohnmayer

unread,
Oct 12, 2014, 6:31:45 PM10/12/14
to electio...@googlegroups.com
In what way would it change the strategic calculus? I would think it would have the effect of making the strategic voter more honest.

Warren D Smith

unread,
Oct 12, 2014, 6:31:51 PM10/12/14
to electio...@googlegroups.com
well, one of the big beefs I have with IRV is it tends to yield
2-party domination. However, plain (non-instant) runoff apparently
does not. One reason for the difference is strategic voter behavior
-- the non-instant voters will always be honest in round #2, even if
strategic.
I would worry that this range+instantrunoff system would have the same
problem and still yield 2-party domination. What says the "NESD
property"?

Clay Shentrup

unread,
Oct 12, 2014, 8:10:00 PM10/12/14
to electio...@googlegroups.com
NESD (Naive Exaggeration Equals Duopoly) happens because of frontrunner polarization. E.g. you prefer Green, Democrat, Republican, and you bump up the Democrat to 1st place to help him defeat the Republican.

With R-IRV, that looks something like changing a sincere vote of
Green=5, Dem=3, GOP=1
to a tactical vote of
Green=5, Dem=5, GOP=0

So in that sense, it's not duopoly-prone.

Now, one might worry that helping the Green into the runoff against the Republican could be dangerous, because the Green would be much less likely than the Democrat to defeat the Republican. With a Plurality-based system, that's a reasonable concern, because a weak candidate might advance purely due to vote splitting. But Score Voting is very different. A candidate who was second, in terms of total score, would presumably be more likely to defeat the #1 scored candidate in a head-to-head race than would be the candidate who was third by total score.

In other words, if Republican=800 points, Green= 701 points, and Democrat = 700 points, do you really want to help the Democrat overtake the Green to make it to the runoff? Do you really think the Democrat is so much more likely than the Green to defeat the Republican, if the total score is really that close?

Additionally, your ability to discriminate between the top two is itself a huge component of your voting power. If you really think the Green is so likely to make it to the runoff that you need to engage in tactics to prevent that, then there's a very good chance that your attempts to block him (i.e. by rating the Green lower than the Democrat) will backfire and help the Dem to defeat the Green in the runoff.

So, actually, it seems that this system may be more resistant to tactical voting.

Warren D Smith

unread,
Oct 12, 2014, 8:11:41 PM10/12/14
to electio...@googlegroups.com
On 10/12/14, Warren D Smith <warre...@gmail.com> wrote:
> well, one of the big beefs I have with IRV is it tends to yield
> 2-party domination. However, plain (non-instant) runoff apparently
> does not. One reason for the difference is strategic voter behavior
> -- the non-instant voters will always be honest in round #2, even if
> strategic.
> I would worry that this range+instantrunoff system would have the same
> problem and still yield 2-party domination. What says the "NESD
> property"?

--Looking into this last question:
If all voters exaggerate A & B to max or min scores (where A & B are
the two major party candidates), then with plain range voting, this does not at
all force either A or B to be the winner. Nader can still easily win.

In contrast, with IRV, such NES voter behavior ("naive exaggeration strategy")
forces either A or B to win.
This is postulated (by me) to be a big reason IRV leads to 2-party
domination, whereas,
with plain range voting, it is hoped (by me) that 2-party domination
can be escaped.

With the range+GenuineRunoff system, what happens?
The voters by thus exaggerating still can elect C & D to the runoff
round, where these
do not necessarily coincide at all with A & B. Also, they could
elect, say, A & C
to the runoff round. At that point voters will vote honestly in the
genuine runoff, so that, again, C could easily win, nothing structural
inherently forces C to lose.

With the range+InstantRunoff system, what happens?
The voters by can elect C & D to the runoff round, where these
do not necessarily coincide at all with A & B. Also, they could
elect, say, A & C
to the runoff round. At that point voters will vote the same way in the runoff
they did in round #1, except now unweighted.
So, it then is possible for C to win, but it sees to me less likely than
in range+GenuineRunoff. The problem is the NES distortion is carried thru into
the InstantRunoff round without any change. With GenuineRunoff, the
NES distortion
is erased in the runoff because the voters will vote honestly in round #2, thus
genuinely electing the better of the two finalists.

So in summary:
my NES theory of 2-party domination does allow range+InstantRunoff to
potentially
escape the 2-party domination trap, which is better than you can say for IRV,
but it still feels less likely to work than plain range and range+GenuineRunoff.

Also, the whole point of the final InstantRunoff round does not appeal to me.
See, with range+GenuineRunoff, the 2nd round really has a point: it is
there to erase strategic distortions in round #1, by now getting
honest X vs Y votes in the final round.
With range+InstantRunoff, there is no such real point to the 2nd
round, it is wholy
and completely an illusory propaganda move motivated by the purpose of
"refuting" an argument which you yourself (Clay S.) have repeatedly
argued does not require refuting.
It actually makes range voting perform worse and adds more complexity. It's
a bit like 3-card monte.

Now finally, as for the claim range voting is not getting anywhere in
the political arena,
so we need something better -- I don't really buy that. See, we have
tolerably good evidence nowadays, from polling studies, that plain
range voting really is wanted by Joe Voter, and Joe would if given the
chance, enact it. Evidence summarized here:
http://ScoreVoting.net/WhatVotersWant.html

It actually seems to be the most-wanted voting system among those we
have such evidence about. So the reason range voting is not getting
anywhere in the political arena
is NOT, far as I can tell, because it is a bad voting system that
needs to be fixed.
It is instead for other reasons:

1. you suck. I suck.

2. where's our multi $M budget?

3. used to get lots of media attention etc?

So my advice is to stay focused on the ACTUAL obstacles that have prevented
success, not invent new fake obstacles. It is possible some other
voting system is better than range voting, and/or more enactible, but
such systems are fairly hard to find
and I would suspect this particular attempt is both a worse system and
less enactible (since more complicated).. It also is desirable to get
more and better evidence about how enactible range & other systems
are. But for the moment, the evidence is not
all that horrible.

Mark Frohnmayer

unread,
Oct 12, 2014, 10:16:49 PM10/12/14
to electio...@googlegroups.com
Hmm.

So if I'm understanding you correctly, the value in the real runoff vs. the instant runoff in your mind is that if it is between candidates that, for strategic reasons, the voter has scored equally in the first round, that there is an honest choice in the second round? When would a strategic score voter reorder his preferences between rounds? My thought is that having an instant top two would up-bias the voter's second choice, where having no runoff at all down-biases the voter's second choice in a strategic calculation. If anything, the "instant" aspect corrects for this up-biasing, because it incentivizes the voter not to "tie" the first and second choices.

I think you're undervaluing the potential for a consensus solution between various election reform advocates. The back and for the between CES and FairVote over IRV vs. Approval was not helpful in getting the unified primary advanced in Oregon, whereas if it were possible to construct a solution that proponents of IRV and proponents of cardinal systems both were reasonably happy with, it would make reform easier.


--
You received this message because you are subscribed to a topic in the Google Groups "The Center for Election Science" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/electionscience/JK82EFn7nrs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to electionscien...@googlegroups.com.

Warren D Smith

unread,
Oct 13, 2014, 3:37:22 PM10/13/14
to electio...@googlegroups.com
> So if I'm understanding you correctly, the value in the real runoff vs. the
> instant runoff in your mind is that if it is between candidates that, for
> strategic reasons, the voter has scored equally in the first round, that
> there is an honest choice in the second round? When would a strategic score
> voter reorder his preferences between rounds?

--it is rare, but it can happen, for it to be strategic to misorder
with score voting:
http://rangevoting.org/RVstrat2.html

--But mostly I would expect voters to dishonestly score the Dem maximum and
the Republican minimum (or vice versa) to try to increase impact of
their vote.
This kind of NES (naive exaggeration strategy)
is very common for Australian rank-order voters (observed fact, not
just hypothesis;
about 80-90% of Australian ballots are in this style),
and it causes distortion which
causes worse winner on average than if all voters had been 100% honest.
(Which is confirmed by computer sims, not just a hypothesis.)

The point of adding a top-2 genuine runoff 2nd round is that voters
will always be honest in that 2nd round, thus hopefully correcting for
this distortion in the 1st round (which involves much dishonest
voting). Hopefully, this correction
effect would give us more gains than losses.

But if you instead add a fake 2nd round (really just regurgitating the
votes from round #1 "instantly") then the distortion is just re-used
with no corrective effect! Pointless!

Now with IRV (instant runoff voting) this whole effect in Australia
causes massive 2-party domination in IRV seats. (Observed fact. Only
one 3rd party candidate managed to win an IRV seat in their house,
just once, during the last 600 races.)
Meanwhile in plurality+top2runoff (genuine runoff) countries, one
observes that there usually is NOT 2-party domination. That is one
big, and real, difference between genuine versus instant runoff:
http://www.rangevoting.org/TTRvsIRVrevdata.html
http://www.rangevoting.org/TTRvIRVstats.html

Now: for the proposed "instant score off"
system IF we permit continuum scores,
it might not be this bad, because Nader voters who understood all that could
vote in round #1 like this:

NADER=9
GORE=8.99999
BUSH=0

where here this voter had dishonestly exaggerated Gore & Bush.
This ballot would still offer honest order for use in the fake 2nd round, thus
allowing Nader to win,
but also would offer almost-maximal strategic exaggeration re Gore & Bush.

In order to do this, though, you'd need continuum scores, or at least
very fine grained (0-9 not good enough, need 0-99 or 0-999) and you'd
need almost all voter to appreciate this.
The former would be less practical and the latter might be unrealistic.

Note that this "8.99999 option" is not available with IRV or
greatest-median-wins versions
of Score, but is available with ordinary sum-of-scores-based score.
My point is we could at least hope that ordinary score voting could survive
the "instant score off" modification without forcing 2-party domination and
re-used-distortion pathologies. But this is only a hope, and maybe
in real life a vain one.

A different way to view the "instant" 2nd round which sounds a lot worse
to me than Clay Shentrup's view is:
"Find the top two A & B.
If A beats B on scores, but B is preferred over A by majority, then B wins."

So we are artificially replacing the right winner with the wrong
winner in this circumstance.
Old-style score voting had the benefit that in a "tyranny of majority
situation" such as a vote on whether to "enslave blacks," the minority
(with honest voting) could still win.
Almost no voting system besides score enjoys this benefit.
With the instant 2nd round added, that unique benefit is sacrificed.

P.S.
Poundstone made some mistakes in his book, some of which he has agreed
with me about in hindsight. For example, see
http://www.rangevoting.org/IrvParadoxProbabilities.html
quote by Poundstone near end of page.
In his book, the FairVote IRV-propagandists told Poundstone (quoted in his book)
the probability of "paradoxes" using IRV was "comparable to a major
meteorite strike" and "has never happened" in practice. In fact it
has already happened in IRV elections brought to us by FairVote
themselves, and it is extremely common, as this page shows.
Poundstone in retrospect wishes he had refuted those bogus FairVote claims, but
they succeeded in conning him at the time.

Warren D Smith

unread,
Oct 13, 2014, 3:44:31 PM10/13/14
to electio...@googlegroups.com
> P.S.
> Poundstone made some mistakes in his book, some of which he has agreed
> with me about in hindsight. For example, see
> http://www.rangevoting.org/IrvParadoxProbabilities.html
> quote by Poundstone near end of page.
> In his book, the FairVote IRV-propagandists told Poundstone (quoted in his
> book)
> the probability of "paradoxes" using IRV was "comparable to a major
> meteorite strike" and "has never happened" in practice. In fact it
> has already happened in IRV elections brought to us by FairVote
> themselves, and it is extremely common, as this page shows.
> Poundstone in retrospect wishes he had refuted those bogus FairVote claims,
> but
> they succeeded in conning him at the time.

--Here I should have been a bit kinder to Poundstone: they only partly
succeeded in conning him. And he did in his book give at least one
historical example of a genuine nonmonotone election, so he already
knew FairVote's claim that had "never" happened was bull.

Mark Frohnmayer

unread,
Oct 13, 2014, 3:53:52 PM10/13/14
to electio...@googlegroups.com
--But mostly I would expect voters to dishonestly score the Dem maximum and
the Republican minimum (or vice versa) to try to increase impact of
their vote.
This kind of NES (naive exaggeration strategy)
is very common for Australian rank-order voters (observed fact, not
just hypothesis;
about 80-90% of Australian ballots are in this style),
and it causes distortion which
causes worse winner on average than if all voters had been 100% honest.
(Which is confirmed by computer sims, not just a hypothesis.)

If anything, this proves the point that the "instant" top two would improve on a one-round system, if we expect there to be strategic voters. Whether the strategic voter exaggerates support for the 2nd choice or "bullet votes" and minimizes support for the second choice, the knowledge that there will be a runoff will make the voter balance the exaggeration dishonesty with the desire to express preference ordering.
 

The point of adding a top-2 genuine runoff 2nd round is that voters
will always be honest in that 2nd round, thus hopefully correcting for
this distortion in the 1st round (which involves much dishonest
voting).  Hopefully, this correction
effect would give us more gains than losses.

But if you instead add a fake 2nd round (really just regurgitating the
votes from round #1 "instantly") then the distortion is just re-used
with no corrective effect! Pointless!

See above.
 
Now: for the proposed "instant score off"
system IF we permit continuum scores,
it might not be this bad, because Nader voters who understood all that could
vote in round #1 like this:

NADER=9
GORE=8.99999
BUSH=0

If we're trying to remove distortion, less resolution would actually be favorable.
 
A different way to view the "instant" 2nd round which sounds a lot worse
to me than Clay Shentrup's view is:
   "Find the top two A & B.
If A beats B on scores, but B is preferred over A by majority, then B wins."

So we are artificially replacing the right winner with the wrong
winner in this circumstance.
Old-style score voting had the benefit that in a "tyranny of majority
situation" such as a vote on whether to "enslave blacks," the minority
(with honest voting) could still win.
Almost no voting system besides score enjoys this benefit.
With the instant 2nd round added, that unique benefit is sacrificed.


Once scoring is in place politically, it's pretty easy to change the computation formula if it looks like it would be preferable.

Warren D Smith

unread,
Oct 13, 2014, 5:18:21 PM10/13/14
to electio...@googlegroups.com
On 10/13/14, Mark Frohnmayer <mark.fr...@gmail.com> wrote:
>> --But mostly I would expect voters to dishonestly score the Dem maximum
>> and
>> the Republican minimum (or vice versa) to try to increase impact of
>> their vote.
>> This kind of NES (naive exaggeration strategy)
>> is very common for Australian rank-order voters (observed fact, not
>> just hypothesis;
>> about 80-90% of Australian ballots are in this style),
>> and it causes distortion which
>> causes worse winner on average than if all voters had been 100% honest.
>> (Which is confirmed by computer sims, not just a hypothesis.)
>>
>
> If anything, this proves the point that the "instant" top two would improve
> on a one-round system, if we expect there to be strategic voters.

--not necessarily so at all.

Let me give concrete example. Let's say with an 0-9 scale, that
NADER is the best winner with honest average score 6. Meanwhile all
voters honestly regard GORE
and BUSH as 2 to 8, but exaggerate them to 0 and 9. Meanwhile assume NADER is
left honestly scored. If neither GORE or BUSH is
preferred over the other by a 2:1 margin or more, then
the (good) result with plain score voting is:
NADER wins with 6, while BUSH and GORE both have below 6, as mean scores.

With instant score-off, say NADER & BUSH are the 2 finalists, then
BUSH automatically
wins the final thanks to the exaggerating of BUSH scores by a voter majority.
E.g. here are 3 fully explicit examples of this or related stuff

#voters their honest vote their exagg vote
51 B=7 N=6 G=2 B=9 N=6 G=0
49 G=7 N=6 B=2 G=9 N=6 B=0
plain score N wins N wins
score+instant NB, then B wins NB then B wins

#voters their honest vote their exagg vote
26 B=7 N=6 G=2 B=9 N=6 G=0
25 B=7 N=5 G=2 B=9 N=5 G=0
25 G=7 N=7 B=2 G=9 N=9 B=0
24 G=7 N=6 B=2 G=9 N=6 B=0
plain score N wins N wins
score+instant NB, then B wins NB then B wins

#voters their honest vote their exagg vote
49 B=7 N=6 G=2 B=9 N=6 G=0
3 N=8 B=7 G=2 B=9 N=9 G=0
23 G=7 N=8 B=2 G=9 N=9 B=0
23 G=7 N=5 B=2 G=9 N=5 B=0
2 G=7 N=4 B=2 G=9 N=4 B=0
plain score N wins N wins
score+instant NB, then N wins NB then B wins

In all these examples NADER was the "best" winner using honest score
voting, also with dishonest-exagg votes despite the distortion NADER still wins
with plain score; but Nader is denied by score+instant.
In the last of these 3 elections, N would win with score+instant if
the voters in the 2nd line exaggerated BUSH to only 8.9999.

If you do not love NADER, just realize this same pathology would
happen with *any*
2 major parties plus one (overall better) 3rd party candidate
appealing to all sides.
So we might very justifiably worry that the score+instant system would
yield 2-party domination. And I think this scenario is not artificial,
it is quite realistic.

Pretty damning in my view.

In case you are wondering, 2-party domination is pretty nearly a death
blow for democracy, removing the "market of choices," making big money
domination and corruption much easier, and removing entire classes of
ideas from even being able to be
discussed at all in the legislature -- and renders improved voting
systems largely irrelevant.

> Whether
> the strategic voter exaggerates support for the 2nd choice or "bullet
> votes" and minimizes support for the second choice, the knowledge that
> there will be a runoff will make the voter balance the exaggeration
> dishonesty with the desire to express preference ordering.

--A fine fantasy. Reality is the Australian ballot behavior, however
which shows that
Australian voters, 80-90% of the time, vote one of the two major
parties top and the other bottom or 2nd bottom out of about 7 parties
on ballot each time. Despite
their "knowledge" that this will yiel 2-party domination, they do it anyway,

Similarly, despite in the USA, voter's "knowledge" that by refusing to
vote 3rd party, they will et 2-party domination, they do so anyway.
Wake up: some systems react badly to common strategies, and I refuse
to accept some pollyanna fantasy that voters will realize those
strategies lead to bad effects hence not use them. That is a fantasy
which has
been disproven over and over in 100+ years of real world experience.
That is the major reason we NEED better voting systems!



>> NADER=9
>> GORE=8.99999
>> BUSH=0
>>
>
> If we're trying to remove distortion, less resolution would actually be
> favorable.

--huh?


> Once scoring is in place politically, it's pretty easy to change the
> computation formula if it looks like it would be preferable.

--I do not know what that meant. But I think if you have a system
whose rules keep getting tweaked by the politicians to make sure they
keep getting elected, that is a disaster just like gerrymandering. If
you or anybody therefore are actually RECOMMENDING that as the right
course of action (?!), BEWARE!
I personally would be EXTREMELY wary of any voting "reformer" who
advocated such tweakage.

--Plain score voting is simpler than instant+score. It works better.
Plain score also is simpler than top2runoff+score (genuine 2nd round).
However in
the latter case, which one is better is not so clear; it depends on
the fractions of voters who are honest & strategic. Adding a
top2runoff can yield better Bayesian Regret than
plain score voting, if there are enough exaggerators in the voter
population, but the
fraction of exaggerators needed is fairly high and the gains are fairly small.

Mark Frohnmayer

unread,
Oct 13, 2014, 6:08:40 PM10/13/14
to electio...@googlegroups.com
Let me give  concrete example.  Let's say with an 0-9 scale, that
NADER is the best winner with honest average score 6.  Meanwhile all
voters honestly regard GORE
and BUSH as 2 to 8, but exaggerate them to 0 and 9.  Meanwhile assume NADER is
left honestly scored. If neither GORE or BUSH is
preferred over the other by a 2:1 margin or more, then
the (good) result with plain score voting is:
   NADER wins with 6, while BUSH and GORE both have below 6, as mean scores.

With instant score-off, say NADER & BUSH are the 2 finalists, then
BUSH automatically
wins the final thanks to the exaggerating of BUSH scores by a voter majority.
E.g. here are 3 fully explicit examples of this or related stuff
...


In all these examples NADER was the "best" winner using honest score
voting, also with dishonest-exagg votes despite the distortion NADER still wins
with plain score; but Nader is denied by score+instant.
In the last of these 3 elections, N would win with score+instant if
the voters in the 2nd line exaggerated BUSH to only 8.9999.

In the first two examples you have Bush preferred by a majority over any other candidate. Sticking to score voting versus an alternative in that scenario may win in theory, but will have a tough time winning in our electorate's concept of democracy. It is a huge advantage for adoption that score+instant elects the majority winner in those cases.

In the third scenario, those 3 voters have sacrificed their ability to differentiate Bush and Nader by over exaggerating Bush. Not a very smart strategy if Nader was their favored candidate.
 
 

If you do not love NADER, just realize this same pathology would
happen with *any*
2 major parties plus one (overall better) 3rd party candidate
appealing to all sides.
So we might very justifiably worry that the score+instant system would
yield 2-party domination. And I think this scenario is not artificial,
it is quite realistic.

Pretty damning in my view.

I'd be very interested to see how it simulates across many election permutations vs. other systems you've simulated.
 

In case you are wondering, 2-party domination is pretty nearly a death
blow for democracy, removing the "market of choices," making big money
domination and corruption much easier, and removing entire classes of
ideas from even being able to be
discussed at all in the legislature -- and renders improved voting
systems largely irrelevant.

I was wondering about that whole 2-party domination thing!
 
--A fine fantasy.  Reality is the Australian ballot behavior, however
which shows that
Australian voters, 80-90% of the time, vote one of the two major
parties top and the other bottom or 2nd bottom out of about 7 parties
on ballot each time.   Despite
their "knowledge" that this will yiel 2-party domination, they do it anyway,

Is this in IRV races, STV races or both? Is it possible that the favorite betrayal in IRV elections affects the strategy for STV elections?
 

Similarly, despite in the USA, voter's "knowledge" that by refusing to
vote 3rd party, they will et 2-party domination, they do so anyway.
Wake up:  some systems react badly to common strategies, and I refuse
to accept some pollyanna fantasy that voters will realize those
strategies lead to bad effects hence not use them. That is a fantasy
which has
been disproven over and over in 100+ years of real world experience.
That is the major reason we NEED better voting systems!

Again, plurality voting punishes honest support of a favorite candidate so yields two party domination. How does this apply to score+instant?

 



>> NADER=9
>> GORE=8.99999
>> BUSH=0
>>
>
> If we're trying to remove distortion, less resolution would actually be
> favorable.

--huh?

The voter has an incentive to express favoritism as well as score. In IRV if I rank Nader 1st, my second choice scores zero in the first round and may get knocked out before I can help him with my vote. In score+runoff, I want to differentiate between my favorite and my second favorite, so I'll rate my favorite as the highest score, and my second choice as highest score - 1. A lower resolution score yields a greater delta between favorite and second favorite.
 


> Once scoring is in place politically, it's pretty easy to change the
> computation formula if it looks like it would be preferable.

--I do not know what that meant.   But I think if you have a system
whose rules  keep getting tweaked by the politicians to make sure they
keep getting elected, that is a disaster just like gerrymandering.  If
you or anybody therefore are actually RECOMMENDING that as the right
course of action (?!), BEWARE!
I personally would be EXTREMELY wary of any voting "reformer" who
advocated such tweakage.

Adoption of reform can happen in steps or all at once. I'm looking for systems that are better on the great Warren Smith List of Systems According To Actual Simulation than the system we have today which is off the bottom of the chart.
 

--Plain score voting is simpler than instant+score.  It works better.

Depends on your definition of working better. Failing to elect the candidate a majority prefer has been  a huge criticism of score voting. This should basically fix that criticism.
 
Plain score also is simpler than top2runoff+score (genuine 2nd round).
However in
the latter case, which one is better is not so clear; it depends on
the fractions of voters who are honest & strategic.   Adding a
top2runoff can yield better Bayesian Regret than
plain score voting, if there are enough exaggerators in the voter
population, but the
fraction of exaggerators needed is fairly high and the gains are fairly small.

Again, it'd be useful to see how it performs given various mixes of honest and strategic voters.

Rob Wilson

unread,
Oct 14, 2014, 4:25:33 AM10/14/14
to electio...@googlegroups.com
The problem I have with this is that I think the ballot will confuse a significant portion of the people.  Too many people misunderstand a simple regular ranking.

Mark Frohnmayer

unread,
Oct 14, 2014, 8:01:27 AM10/14/14
to electio...@googlegroups.com

The ballot looks exactly the same as a normal score ballot. How would people confuse that as a ranking?

On Oct 14, 2014 1:25 AM, "Rob Wilson" <blahf...@gmail.com> wrote:
The problem I have with this is that I think the ballot will confuse a significant portion of the people.  Too many people misunderstand a simple regular ranking.

--

Blahface

unread,
Oct 14, 2014, 2:19:58 PM10/14/14
to electio...@googlegroups.com
I didn't read it carefully.  I assumed there was a ranking and a rating.

Warren D Smith

unread,
Oct 14, 2014, 4:45:30 PM10/14/14
to electio...@googlegroups.com
On 10/13/14, Mark Frohnmayer <mark.fr...@gmail.com> wrote:
--you're wrong according to the poll data here
http://rangevoting.org/WhatVotersWant.html ;
despite the "disadvantage" that score can go against a majority
preference (which actually is an advantage) the fact is, people want
score and would enact it, and it seems more enactible based on this
data than any rival voting system. If you disagree, find poll data
saying the contrary, not just your barefaced assertion. Your
barefaced assertion conceivably still correct even though it goes
against my poll data because as far as I have been able to tell, the
ultimate poll study on this crucial question has not been done; the
polls that were done all are subject to criticism to some extent.
Still, based on the data I've got, plain score voting is highly
enactible and your worries are wrong.

> In the third scenario, those 3 voters have sacrificed their ability to
> differentiate Bush and Nader by over exaggerating Bush. Not a very smart
> strategy if Nader was their favored candidate.

--they exaggerated Nader also to max. The result of both
exaggerations was to pretend Nader=Bush=max.

Now despite your saying "not a very smart strategy" what seems to have
difficulty penetrating is the fact that real voters in Australia use
exactly this strategy 80-90% of the time. Are you interested in the
real world, or in a fantasy world in which voters act like you claim
to act? I mean, this Australian fact is massive, overwhelming
evidence
for the commonness and importance of naive exaggeration strategy,
which in their case is not merely pretending Nader=Bush, but actually
pretending Bush>Nader since rank-equalities for them are forbidden.

Similarly in the USA, actual real polls of actual real voters found
that 90% of those who honestly thought Nader was the best, voted for
one of his rivals. Again, this was not merely pretending that
Nader=Bush, but pretending Bush>Nader.
I repeat. 90%. This is one of the most massive findings in all of
political science.

At some point, massive overwhelming evidence from the real world has
to overwhelm
fantasies about how one wishes things were.

> I'd be very interested to see how it simulates across many election
> permutations vs. other systems you've simulated.

--sigh, well, maybe I'll put it into my simulator, which is available
public source if anybody else wanted to do it:
http://rangevoting.org/IEVS/IEVS.c

I in fact had done sims and had proven a theorem about 2-candidate BRs for
plurality and for range voting, to quantify the disadvantage plurality
suffers even
in the 2-candidate case (with honest voting; with strategic the two
systems become identical).

But you cannot really simulate 2-party domination, a multi-election
effect, without
making assumptions about human behavior, which is something I don't
terribly understand. So my view is, since we do not understand human
behavior in the future,
and 2-party domination is a huge disaster, we should try to play it
safe. There are systems which seem more and less likely to yield 2PD.
Why choose one of the "more"s
when you get no benefits in return, except for a currently completely
unsupported fantasy about enactibility?

> Is this in IRV races, STV races or both? Is it possible that the favorite
> betrayal in IRV elections affects the strategy for STV elections?

--Australia has both STV (Senate) and IRV (house). Voters seem to use
the same strategies in both kinds of voting, probably in part because
the "how to vote" ranking-advice cards issued by Australia's parties
do not distinguish.

The effect that results is: IRV seats become 2-party dominated to a
massive extent. But STV seats (since this is a multiwinner PR system)
are not 2PD in Australia. Thus proving "which voting system" really
really matters about 2PD.

> Again, plurality voting punishes honest support of a favorite candidate so
> yields two party domination. How does this apply to score+instant?

--it applies, but to a lesser extent, due to desires to "maximize impact"
via exaggeration in round #1. If your honest view is Dem=5, Repub=4
on an 0-9 scale, then you are a total idiot if you honestly vote 5 &
4, because your impact is reduced
by a factor 9.

Do you want to be that stupid and ineffectual about the only sub-race
that matters? For a lot of people, their answer will be "no." This
is the case no matter how much you or I or anybody tells them they are
taking a dangerous strategy. (And your advice to them would in any
case be swamped by huge-money loud-blaring advice from the Dem & Repub
parties to "pay no attention to Mark about this.") And then with
score+instant this distortion is re-used and thus amplified by the
pseudo-2nd-round, whereas with plain score it occurs without re-use,
and with score+genuine runoff this distortion actually is reduced.

I do not know if score+instant would yield 2PD, it just seems a
greater risk than with
plain score voting or score+genuine runoff.

>> > If we're trying to remove distortion, less resolution would actually be
>> > favorable.
>>
>> --huh?
>
> The voter has an incentive to express favoritism as well as score. In IRV
> if I rank Nader 1st, my second choice scores zero in the first round and
> may get knocked out before I can help him with my vote. In score+runoff, I
> want to differentiate between my favorite and my second favorite, so I'll
> rate my favorite as the highest score, and my second choice as highest
> score - 1. A lower resolution score yields a greater delta between favorite
> and second favorite.

--no, you do not terribly want "to differentiate between my favorite
and my second favorite"; you want more to differentiate between
SOCIETY's 1st & 2nd favorite's.
That's the place where the impact is. You want to maximize your vote's
impact on what matters. Hence you vote (say) Bush=9, Gore=0, even if
you honestly feel Nader=9, Bush=5, Gore=4. Now Score+Instant with
continuum scores would allow you to
do Nader=9, Bush=8.9999, Gore=0 getting almost-maximal impact on Bush
vs Gore, while still preferring Nader in round #2. So in principle
Score+instant WITH CONTINUUM SCORES would enable voters to
almost-overcome my whole naive-exagg/2PD objection
to score+instant.

If those voters realized that. They'd need to, since this behavior is
a bit unnatural.


> Adoption of reform can happen in steps or all at once. I'm looking for
> systems that are better on the great Warren Smith List of Systems According
> To Actual Simulation than the system we have today which is off the bottom
> of the chart.

The single-winner best (in my sims) systems among those that seem
simple enough, seem to me to be these:

* plain score voting
* Benham's MCA system, which is the same as a version of "majority
judgment" with 3 score levels. Specifically, voter "Disapproves,"
"Approves" or "2approves" each canddt.
Most-2approvals wins if gets >50%, else regular approval-winner wins
(where now 2approve & approve regarded as same thing).
* score voting with a genuine 2nd round (top 2) added.

Jameson Quinn has been claiming his SODA system might be good or even better,
and it sounded like he likely had some points;
but last I looked he'd kept failing to produce any acceptably simple
description of
the SODA system, I think that is probably possible though, in which case I might
also support SODA. SODA is unusual in that both the voters and the
candidates are involved, not just the voters, because voters can give
parts of their votes to candidates
who then can re-vote with them using "proxy power." That can confer
benefits, but
it also makes it impossible to simulate without, again, more
assumptions about human behavior.

> Depends on your definition of working better. Failing to elect the
> candidate a majority prefer has been a huge criticism of score voting.
> This should basically fix that criticism.

--It isn't a "criticism." It is an "advantage." It improves Bayesian
regret with
honest voters, and leave it the same with strategic voters.

But in any case it is quite rare in real life that range voting elects A while
a majority prefers B. This "thwarted majority" phenomenon seems to
happen considerably more often with IRV than with score voting.
In view of that fact, I find it quite odd that the IRV propagandists
at "FairVote" have the balls to consider this a "criticism" of score
voting.

In my attempts to find a real life example of score voting thwarted
majority, the closest
I ever came was the following election: 2009 Los Angeles Mayor race, see
http://www.rangevoting.org/LAmayors.html
which may have been such an example.
You might want to read that and think who you believe should have been elected.

Warren D Smith

unread,
Oct 14, 2014, 4:46:19 PM10/14/14
to electio...@googlegroups.com
correction, I meant 2001 LA mayor, not 2009.

http://www.rangevoting.org/LAmayors.html

Clay Shentrup

unread,
Oct 14, 2014, 11:39:31 PM10/14/14
to electio...@googlegroups.com
On Tuesday, October 14, 2014 1:25:33 AM UTC-7, Rob Wilson wrote:
The problem I have with this is that I think the ballot will confuse a significant portion of the people.  Too many people misunderstand a simple regular ranking.

Scoring is simpler than ranking. E.g. it's pervasive on the internet (Yelp, Amazon.com, etc.) and you can give the same score to multiple candidates. 

Clay Shentrup

unread,
Oct 15, 2014, 12:04:22 AM10/15/14
to electio...@googlegroups.com
On Tuesday, October 14, 2014 1:45:30 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
> Depends on your definition of working better. Failing to elect the
> candidate a majority prefer has been  a huge criticism of score voting.
> This should basically fix that criticism.

--It isn't a "criticism." It is an "advantage."  It improves Bayesian regret with honest voters, and leave it the same with strategic voters.

It may *actually* be an advantage, but it has been a criticism made by actual human beings (e.g. Steve Chessin of CFER), because they do not understand social choice theory. Steve specifically did not understand how Condorcet cycles prove the group may favor someone other than the Condorcet (or even outright majority) winner, because he doesn't understand how IoIA is a requirement of voting (he erroneously thinks of it as a subjective desideratum).

And he went to MIT. Gulp.

It's important to understand what system is actually best. And it's nice to know what voters would pick if it were put to a ballot initiative. But to even getting to the point where a city council member would put your idea on the ballot, or a rich person would give you money to obtain the signatures to do it, requires a huge amount of convincing policy makers, who (in my experience) tend to have the misguided notion that you simply cannot violate "majority rule". Voting theory is just extremely counter-intuitive, and people don't easily understand how "majority rule" can be wrong. It was a huge stumbling block for me as well, until you described Bayesian Regret to me, and I came to understand how voting is (like individual choice) just a matter of maximizing one's personal expected utility. But even for me, huge champion of Score Voting, that took a lot of debate (like, a couple weeks of back and forth with you via email).

I hope you would agree that this proposal would be radically better than IRV. I also hope you can see that it retains certain properties of IRV that unsophisticated people like Rob Richie think are important, plausibly giving it more viability in the election reform space—regardless of what your polls say about what voters would choose.

On that note, think about those polls that showed IRV being so unpopular in San Francisco. Despite that, note that the voters have not actually been given that option. There was serious debate about it a couple years back in San Francisco, but it was quashed because the board of supervisors couldn't get the votes to put it on the ballot. And local progressive leaders (you have to basically talk about how "progressive" you are around here in order to get elected) bristle at the notion of being anti-RCV. After all, RCV improves turnout, and diversity, and civility. Facts which they know to be true despite not knowing monotonicity from a hole in the ground.

But in any case it is quite rare in real life that range voting elects A while
a majority prefers B.   This "thwarted majority" phenomenon seems to
happen considerably more often with IRV than with score voting.
In view of that fact, I find it quite odd that the IRV propagandists
at "FairVote" have the balls to consider this a "criticism" of score
voting.

Well, their argument is that it even makes it possible for a majority-favored candidate to lose, whereas IRV doesn't.

In my attempts to find a real life example of score voting thwarted
majority, the closest
I ever came was the following election: 2009 Los Angeles Mayor race, see
   http://www.rangevoting.org/LAmayors.html

So you're saying that electing the majority-favored between the top two highest-scored candidates is unlikely to elect the second-highest-scored candidate? Great!

Clay Shentrup

unread,
Oct 15, 2014, 12:05:14 AM10/15/14
to electio...@googlegroups.com
On Tuesday, October 14, 2014 11:19:58 AM UTC-7, Blahface wrote:
I didn't read it carefully.  I assumed there was a ranking and a rating.

No. If you score Nader a 3 and Bush a 1, you "ranked" Nader ahead of Bush.

Mark Frohnmayer

unread,
Oct 15, 2014, 2:20:45 AM10/15/14
to electio...@googlegroups.com
On Tue, Oct 14, 2014 at 1:45 PM, Warren D Smith <warre...@gmail.com> wrote:
--you're wrong according to the poll data here
    http://rangevoting.org/WhatVotersWant.html ;
despite the "disadvantage" that score can go against a majority
preference (which actually is an advantage) the fact is, people want
score and would enact it, and it seems more enactible based on this
data than any rival voting system.  If you disagree, find poll data
saying the contrary, not just your barefaced assertion.  Your
barefaced assertion conceivably still correct even though it goes
against my poll data because as far as I have been able to tell, the
ultimate poll study on this crucial question has not been done; the
polls that were done all are subject to criticism to some extent.
Still, based on the data I've got, plain score voting is highly
enactible and your worries are wrong.

There's polling and then there's polling. I've been working on getting a measure passed in Oregon for the last year - the basic concept polled 72-18, but it'll be lucky if the measure passes at all. Opposition can and will sow FUD, and they'll poll test messaging "It's un-American to have the preferred candidate of the major it be beaten by math!" and then just hammer on it until it sinks the measure. Having consensus in the reform community is valuable to soften this onslaught. If there ever were a system that everyone could agree is "worth a try", relaxing some hardened positions would be worth it.
 

--they exaggerated Nader also to max.  The result of both
exaggerations was to pretend Nader=Bush=max.

So worst case is that they approval vote?
 

Now despite your saying "not a very smart strategy" what seems to have
difficulty penetrating is the fact that real voters in Australia use
exactly this strategy 80-90% of the time.  Are you interested in the
real world, or in a fantasy world in which voters act like you claim
to act?  I mean, this Australian fact is massive, overwhelming
evidence
for the commonness and importance of naive exaggeration strategy,
which in their case is not merely pretending Nader=Bush, but actually
pretending Bush>Nader since rank-equalities for them are forbidden.

Australia is a classic example of favorite betrayal producing strategic voting. See: 


"At the 2007 federal election progressive people were desperate to kick out the conservative Liberal Party government, even though most Greens voters had little illusions about Labor leader Kevin Rudd. He certainly didn't have the appeal of an Obama-like figure. However a lot of very politically aware people told me that they would be voting Labor '1' because they didn't want to risk helping the Liberals by splitting the vote. This is despite the fact that a '1' vote for the Greens and a '2' vote for Labor would have been just as valuable in defeating a Liberal candidate. This confusion is often encouraged by the major parties who do not want people to give a first preference to a minor party. In left-wing inner-city seats around Sydney and Melbourne, where the Greens are now challenging the hold of the Labor Party, Labor campaigners often will claim that a vote for the Greens would help the Liberal Party, sowing confusion about our electoral system, in order to bring progressive voters back to Labor."

I would think that the use of IRV in some elections would create the misperception that the same strategy ought to be used in STV.
 

Similarly in the USA, actual real polls of actual real voters found
that 90% of those who honestly thought Nader was the best, voted for
one of his rivals.  Again, this was not merely pretending that
Nader=Bush, but pretending Bush>Nader.
I repeat.  90%.  This is one of the most massive findings in all of
political science.

People are well aware of the spoiler effect with plurality voting.
 

At some point, massive overwhelming evidence from the real world has
to overwhelm
fantasies about how one wishes things were.

Voters vote strategically. The case we're talking about above is neither plurality nor IRV, which both have flaws that can punish the voter badly for voting honestly.

Do your simulations measure how much regret is experienced by strategic voters? I.e. can you validate strategies as either being more or less effective for the strategic voter in terms of minimizing his own regret?

I think there will be a lot of voters who will want to express a preference between their first and second choices. That comes from interacting with literally thousands of voters over the course of the year explaining a new system to them.
 

> I'd be very interested to see how it simulates across many election
> permutations vs. other systems you've simulated.

--sigh, well, maybe I'll put it into my simulator, which is available
public source if anybody else wanted to do it:
   http://rangevoting.org/IEVS/IEVS.c


That would be awesome!
 
and 2-party domination is a huge disaster, we should try to play it
safe.  There are systems which seem more and less likely to yield 2PD.
Why choose one of the "more"s
when you get no benefits in return, except for a currently  completely
unsupported fantasy about enactibility?

I don't think you've made anything close to a connection substantiating that this system is likely to produce 2PD.


> Again, plurality voting punishes honest support of a favorite candidate so
> yields two party domination. How does this apply to score+instant?

--it applies, but to a lesser extent, due to desires to "maximize impact"
via exaggeration in round #1.  If your honest view is Dem=5, Repub=4
on an 0-9 scale, then you are a total idiot if you honestly vote 5 &
4, because your impact is reduced
by a factor 9.
 
I would expect voters to always weight their range between the min and max value. But if I know that two advance to a simple preference test, I have a reason to rate two highly.
 

Do you want to be that stupid and ineffectual about the only sub-race
that matters?  For a lot of people, their answer will be "no."  This
is the case no matter how much you or I or anybody tells them they are
taking a dangerous strategy.   (And your advice to them would in any
case be swamped by huge-money loud-blaring advice from the Dem & Repub
parties to "pay no attention to Mark about this.")   And then with
score+instant this distortion is re-used and thus amplified by the
pseudo-2nd-round, whereas with plain score it occurs without re-use,
and with score+genuine runoff this distortion actually is reduced.

I think a lot of voters will not willingly sacrifice the ability to express true favoritism (i.e. expressing highest score for their favorite, highest score - 1 for their acceptable front runner). This inability to differentiate is at the core of a big chunk of the criticism of Approval Voting. The only time I wouldn't express a preference order is when I am convinced that the race is super close between the two frontrunners.
 

I do not know if score+instant would yield 2PD, it just seems a
greater risk than with
plain score voting or score+genuine runoff.

Hmm... Not totally seeing it.
 

>> > If we're trying to remove distortion, less resolution would actually be
>> > favorable.
>>
>> --huh?
>
> The voter has an incentive to express favoritism as well as score. In IRV
> if I rank Nader 1st, my second choice scores zero in the first round and
> may get knocked out before I can help him with my vote. In score+runoff, I
> want to differentiate between my favorite and my second favorite, so I'll
> rate my favorite as the highest score, and my second choice as highest
> score - 1. A lower resolution score yields a greater delta between favorite
> and second favorite.

--no, you do not terribly want "to differentiate between my favorite
and my second favorite"; you want more to differentiate between
SOCIETY's 1st & 2nd favorite's.
That's the place where the impact is. You want to maximize your vote's
impact on what matters. Hence you vote (say) Bush=9, Gore=0, even if
you honestly feel Nader=9, Bush=5, Gore=4.  Now Score+Instant with
continuum scores would allow you to
do Nader=9, Bush=8.9999, Gore=0  getting almost-maximal impact on Bush
vs Gore, while still preferring Nader in round #2.  So in principle
Score+instant WITH CONTINUUM SCORES would enable voters to
almost-overcome my whole naive-exagg/2PD objection
to score+instant.

I think you have it backwards. As a voter I very much want to be able to express preference as well as approval, and unless the race is super tight between the frontrunners, I won't see it as very risky to score my frontrunner with top score - 1 in order that if my actual favorite makes it to the runoff that he or she gets my full support. If I top score (or bottom score) multiple candidates, I give up the ability to express preference in the runoff round. So I won't do full naive exaggeration.

I can barely keep my eyes open, so if that stuff didn't make sense I'll revise it in the morning.

Cheers,
Mark

Warren D Smith

unread,
Oct 15, 2014, 1:38:51 PM10/15/14
to electio...@googlegroups.com
> I think you have it backwards. As a voter I very much want to be able to
> express preference as well as approval, and unless the race is super tight
> between the frontrunners, I won't see it as very risky to score my
> frontrunner with top score - 1 in order that if my actual favorite makes it
> to the runoff that he or she gets my full support. If I top score (or
> bottom score) multiple candidates, I give up the ability to express
> preference in the runoff round. So I won't do full naive exaggeration.

--But the Australians do do full exaggeration, 80-90% of time.
Admittedly using different voting systems than this one, but I'm not
convinced they'll
care much what the voting system is. I've never been able to detect
any real-world-voter strategic behavior differences between any two
voting systems with the same kind of ballots. They probably exist
but probably are so small they are hard to measure.

That is why I think "reaction to NES strategy" is a good tool to use
to (1) decide between
voting systems and (2) to worry about 2-party-domination. Because NES
strategy is a very simple GENERAL PURPOSE approach that voters can use
on essentially any voting system at all, without needing to know the
details. It has been proved to be effective on a very wide class of
voting systems, in theory; and in practice the Australian and
USA-Nader-2000 facts I mentioned show it is extremely common.

Next, the entire argument for score+instant seems to be based on
unsupported hypotheses about enactibility and worries about FUD from
opponents.
Well,

**1**
I think the best approach to counter FUD, is to have the SIMPLEST & CLEAREST
AND MOST CONCISELY-DESCRIBED voting system (or whatever it is)
possible. Voter will be least uncertain about something simple.
There have been many many referenda in which the proponents just could
not stop themselves from making it many pages long with many
complicated details. They just could not bring themselves to shut up.
They just had to add yet another gimmick to counteract yet another
illusory flaw, that they knew full well was an illusory flaw. As soon
as voter sees that, she worries many of the details she does not
understand, might be loopholes and gimmicks, and tends to vote against
it. You want simple referenda that fit in half a page or less. And
if you think this is impossible, then I suggest you consider the
actual wordings of actual recent referenda in Britain for IRV,
Scotland independence, etc. If you cannot make your wordings 1 page
or less, then forget about it.

Range voting is simplest. Adding new stuff to it like this makes it
less simple. That creates more, not less, FUD.

2. Yes, opponents can object to the anti-majority property, but they
will be countered by the fact that they are wrong. One could hope
this would neutralize that objection, or better.

Now about IRV (which apparently elects A, despite B>A majority, more
often than range does)... you correctly noted that this is a different
phenomenon than:
X is majority-top rated, but loses.
IRV, plurality, & Condorcet always elect X, but range & Borda can
sometimes fail
to elect X. Well, I think this is a very rare effect for range voting.
I'm unaware of any example in real life (or pseudo-real-life based on
polls about elections where we try to see what would have happened
with other systems). And again I would still contend that if+when it
happens, it is probably a good thing and range was right.

3. Although you worry about opponents complaining re majority -- you
should also worry
about opponents pointing out the fact you are lying. Specifically, if
you (A) contend score is a good thing, more information than just
ranking or plurality, but (B) contend that majority with scores
disregarded and intentionally erased suddenly is the better thing just
for 2 particular candidates, then you are contradicting yourself. An
opponent trying to sow FUD could quite rightly point out "you are
contradicting yourself, therefore we know at least one of these
statements A&B is a lie."

So, watch out.

Now if you want my opinion, the REAL potential problems that opponents
have managed to point out about range voting are not this majority
illusion.
** "1-sided strategy" worries.
** implementation worries related to having to get new voting machines
in some cases.

I think those are the two biggest possibly-genuine objections.
I happen to be unconvinced "one sided strategy" will happen in the
real world, and still have seen no evidence it exists. (I find it
interesting/depressing that Jameson Quinn, a big objector to Range on
this point, who got funding to do actual polling studies and to
subject experimental subjects to actual made-up election scenarios to
see how they'd react, was either unwilling or unable to find 1-sided
strategy going on, and also unwilling/unable to investigate voting
method enactibility. Why stay away from the most important two
questions?)
The second objection will be more or less valid depending on what
machinery they
already have in that state, and should be investigated. But it seems
to me score+instant will always be at least as bad as plain score on
this point.

Warren D Smith

unread,
Oct 15, 2014, 5:07:02 PM10/15/14
to electio...@googlegroups.com
About simulations.

The simulations I already did ages ago, tell you quite a lot. For
100% HONEST voters,
score+instant is the same thing as score+genuine runoff.

For 100% strategic voters... well, it depends on what strategy they
use... if they
use the most naive "exaggerate the 2 frontrunners to max & min, then, all other
candidates you like more then the best frontrunner get max, all you like
less than the worst frontrunner get min" then it still is not clear
what the remaining
candidates should be scored. If you somehow score them all maxes & mins,
then score+instant, score, and approval all should be the same thing.
Meanwhile, score+runoff = approval+runoff will both be equivalent, but these 2
should be different and better versus the preceding group because
their final top2 runoff
will involve honest voting.

In the scenarios simulated here
http://rangevoting.org/StratHonMix.html

100% HONEST BRs (Bayesian Regrets):
MagicBest 0
Plain Range 0.0480
Borda 0.0938
Range+Runoff 0.1206 = Range+Instant
MCA 0.1784
Bucklin 0.2031
IRV 0.2168
Plur+runoff 0.2360
Plain Plurality 0.3314
Random Winner 1
Note, for 100% honest voters, adding a runoff to plain range makes
it perform substantially worse and also makes it more complicated and
requiring 2 rounds.

But for 100% strategic voters, the tables are turned; adding an honest
top2 runoff
improves BR (although the gain is not as large as in the other
direction in the 100% honest case) thanks to the partial
distortion-removal effect. But adding an instant top-2
runoff has no effect because approval and approval+instant are the same thing.

Range+GenuineRunoff tends to improve slightly over plain Range (up to
10-30% regret reduction) when 75% or more of the voters are strategic.
But when 75% or more of the voters are honest, plain range is better
than Range+Runoff by a lot (up to about 3 times smaller regret, as
seen above).

Based on that, while I have not got the numbers right now, they should
be something like this

100% STRATEGIC BRs:
MagicBest 0
Approv+runoff 0.20 same as score+runoff
Plain Approval 0.25 same as score and score+instant
Random Winner 1


But anyhow you can see without any simulation that score+instant
is the same as plain score in the full-strategy case, and worse than plain score
in the full-honesty case. In the case of strat-honest voter mixtures presumably
it always is worse (by "interpolation").

Mark Frohnmayer

unread,
Oct 19, 2014, 3:19:12 PM10/19/14
to electio...@googlegroups.com
Warren,

Are you doing any measure of regret just for the strategic voters in the simulation - i.e. when you test a proposed strategy can you tell if it actually advantages the voters who use it? For example in the case of plurality or IRV, a strategic vote for the lesser evil front runner as a first choice can actually produce a better outcome for those strategic voters than an honest vote would have.

The two proposed theories for strategic range voting are naive exaggeration (basically a strategic approval vote) and bullet voting (an honest plurality vote).

If the voter knows there will be an instant runoff between the two highest rated candidates, this should change the strategic calculus - because giving the same rating to multiple candidates sacrifices the ability to have any determinative power should the election be between those two.

In practice, that means that if I'm a naive exaggerator, instead of giving max_score to the lesser evil and all I like more, I'll give max_score to my actual favorite and max_score - 1 to the lesser evil and anyone else I like more than the lesser evil.

Likewise if I'm a bullet voter, I'll at least give the lesser evil and anyone I like more a rating of min_score + 1.

In both cases the strategic voter is giving up 1/max_score power in the first round in order to gain 100% voting power in the second round, which is a strategically good trade unless the lesser evil and greater evil poll very close and the actual favorite has no chance.

It'd be awesome to see the BR simulations of this method given that proposed strategy.

Cheers,
Mark



William Waugh

unread,
Oct 20, 2014, 10:40:18 PM10/20/14
to electio...@googlegroups.com
What if the ballot contained a score section for the first phase of the tallying, and below it, a ranking section for the top-two phase?  I guess most voters would say, "why do you need all that complexity", and they'd dismiss the idea, and go back to saying "Voter ID".

As for the point that advocates of Score-off could be accused of "lying", maybe they could forestall that by saying "we believe plain Score is slightly better, but we want to compromise with another faction with whom we are not in full agreement, so as to achieve unity around a system we think will work much better than Plurality to deliver nearly equal political power to each citizen."

On Sunday, October 12, 2014 2:32:38 AM UTC-4, Clay Shentrup wrote https://groups.google.com/d/msg/electionscience/JK82EFn7nrs/Lble3V2CW4UJ and there were lots of comments following up.

Mark Frohnmayer

unread,
Oct 20, 2014, 11:13:20 PM10/20/14
to electio...@googlegroups.com


On Oct 20, 2014 7:40 PM, "William Waugh" <google_wil...@spamgourmet.com> wrote:
>
> What if the ballot contained a score section for the first phase of the tallying, and below it, a ranking section for the top-two phase?  I guess most voters would say, "why do you need all that complexity", and they'd dismiss the idea, and go back to saying "Voter ID".

Yes William. The advantage of using scores is that the ranks can be trivially derived from the scores. This idea actually came out of a conversation with Rob Richie at the equal vote conference ... He suggested that approval plus top two could be done on one ballot (for overseas voters and the like) if there was an approval part and a rank order. By using score you get rid of the need for two separate ways of describing your vote.

>
> As for the point that advocates of Score-off could be accused of "lying", maybe they could forestall that by saying "we believe plain Score is slightly better, but we want to compromise with another faction with whom we are not in full agreement, so as to achieve unity around a system we think will work much better than Plurality to deliver nearly equal political power to each citizen."

Well partly it's for consensus, but also I believe the system would discourage strategic score voting because the voter has to sacrifice his ability to express second round preference if he either fully naively exaggerated or fully bullet votes.

Under what conditions that would beat out pure score is a question for the mighty oracles of election simulation (whether or not they think voters will actually vote in that less strategic way).

>
> On Sunday, October 12, 2014 2:32:38 AM UTC-4, Clay Shentrup wrote https://groups.google.com/d/msg/electionscience/JK82EFn7nrs/Lble3V2CW4UJ and there were lots of comments following up.
>

William Waugh

unread,
Oct 21, 2014, 8:37:58 PM10/21/14
to electio...@googlegroups.com
On Monday, October 20, 2014 11:13:20 PM UTC-4, Mark Frohnmayer wrote:


On Oct 20, 2014 7:40 PM, "William Waugh" <google_wil...@spamgourmet.com> wrote:
>
> What if the ballot contained a score section for the first phase of the tallying, and below it, a ranking section for the top-two phase?  I guess most voters would say, "why do you need all that complexity", and they'd dismiss the idea, and go back to saying "Voter ID".

Yes William. The advantage of using scores is that the ranks can be trivially derived from the scores.

They can, but it changes the meaning of the scores.  I am still curious as to what CES members think. Lay aside for the moment the question of getting acceptance.  Could score followed by top-two instant runoff based on a separate ranking section produce a more satisfying result than score alone?

Mark Frohnmayer

unread,
Oct 21, 2014, 8:49:05 PM10/21/14
to electio...@googlegroups.com
It depends on the honesty level of the voters - if you had a scoring and a rank ordering that let you determine a true runoff, that's score2runoff  - does better with strategic voters than any other system on Warren's list. If you had all honest score voters, then score voting would outperform it.

The hypothesis behind the derived rank runoff is that it will cause a more honest scoring strategy to be adopted than either naive exaggeration or bullet voting, but that strategy would have to be coded to know for sure.

--

Warren D Smith

unread,
Oct 22, 2014, 1:48:20 AM10/22/14
to electio...@googlegroups.com
> Yes William. The advantage of using scores is that the ranks can be
> trivially derived from the scores.

--except if equal...

> This idea actually came out of a
> conversation with Rob Richie at the equal vote conference ... He suggested
> that approval plus top two could be done on one ballot (for overseas voters
> and the like) if there was an approval part and a rank order. By using
> score you get rid of the need for two separate ways of describing your
> vote.

--Oh great, it figures a bad idea like this would come from Rob Richie...

>> As for the point that advocates of Score-off could be accused of "lying",
> maybe they could forestall that by saying "we believe plain Score is
> slightly better, but we want to compromise with another faction with whom
> we are not in full agreement, so as to achieve unity around a system we
> think will work much better than Plurality to deliver nearly equal
> political power to each citizen."
>
> Well partly it's for consensus,

--failed at that...

> but also I believe the system would
> discourage strategic score voting because the voter has to sacrifice his
> ability to express second round preference if he either fully naively
> exaggerated or fully bullet votes.

--groan. "I know, let's intentionally make a system that reacts
badly to strategic voting,
because that'll discourage strategic voting!" That plan never worked
in the history of the universe, far as I can tell...

> Under what conditions that would beat out pure score is a question for the
> mighty oracles of election simulation (whether or not they think voters
> will actually vote in that less strategic way).

--arghh. This all is a fix of a non-problem. If it ain't broke, don't fix it.
To understand one reason it was a non-problem, consider
http://rangevoting.org/AppCW.html
(very simple argument offered).

Although it is possible in principle for a score election to
(1) elect A, but B is preferred over A by a voter majority
(2) elect A, but B is ranked top by voter majority
that page explains, with a very very simple argument, why both are
rare in reality due to voters who strategize automatically creating a
force which eliminates both these phenomena.
In fact, theorem is proven that under a certain simple and pretty
realistic voter behavior model (and in the absence of condorcet
cycles), both these problems actually become impossible.

This naturally present force already almost eliminated this so-called
problem. You do not need an ugly, complicated, worse, extra hack to
"fix" this problem. It already is not a problem. You are just busy
inventing non-problems to fix, rather like a chicken with head cut
off. You think this will impress people, but it does not. The notion
voters like it is currently supported by no evidence. There is,
however, genuine poll evidence voters do want plain score voting.

Also IRV suffers (1) more often than score in practice, so (1) simply
is not a valid criticism of score by IRVers.

Also, although IRV does not suffer (2), while score can -- (2) is
EXTREMELY rare. I've never seen an instance where it even came close
to happening, and in the Random Election Model it is exponentially(V)
rare where V=#voters. You should not design voting systems based on
worries about the Loch Ness Monster. You need to design
voting systems based on actually-genuine worries, i.e. problems that
actually arise in
real life experience a noticeable amount.

In a fictional world without such a force (because in said world all
voters are honest and none strategic) see h
http://rangevoting.org/RandElect.html
http://rangevoting.org/ReasAssump.html#app
for some sim data about how often some such phenomena then occur...
answer is, (1) is reasonably common in this sim.

Back in the day, Frohnmayer suggested the (not equivalent to score+instant) idea
of Approval + a second top-2 runoff round. The second round was a
genuine second round, not an instant fake round. That suggestion
already pissed off (author
of book "approval voting") Brams, but after persuasion he held his
nose and supported it. And in fact, this approval+genuineTop2runoff
suggestion was not all that bad. Why?
Well, despite cavalierly ruining practically every nice logical
property enjoyed by approval voting, and making it more complicated,
and making it 2 rounds, it nevertheless does happen to behave well in
computer sims by Fishburn and by me. In fact, a first round with
approval electing the top M, then a 2nd approval voting round among
those M only
to elect 1 final winner, is found by some measures to work BEST if
M=2. Note, better than M=1 (plain approval). And the 2-man final
round has the advantage all its voters will be honest thus
compensating for strategic distortions in round #1 which can happen if
there are too many strategists.

So: that was ok in my view.

But the fake second round idea, is just stupid. It does not correct
for strategic distortions. It does not correct for anything. It in
fact takes valuable score data you'd painstakingly gathered, and
intentionally throws it in the garbage just at the crux
point of deciding the winner, intentionally using worse/less data for
that crux decision.

It also cavalierly sacrifices just about every logical property plain
score enjoys (simplicity, monotonicity, voting is better than not
voting, subdistrict consistency).

And for WHAT do we make these sacrifices?
To "fix" a non-problem raised by a fool with zero evidence?

And wait, there is more. Even when "problems" (1) and/or even the ultra-rare (2)
do occur, they actually on average are not problems: plain score
voting's winner is better
than B on average.

There are always idiots who make stupid criticisms of stuff. The
correct response
is to explain why they are wrong. It is not to say "oh, I see! I too
will now intentionally become an idiot so I can gain the admiration of
other idiots!"

How about the (more common!) criticism "score voting is not plurality voting!!!"
Well, we have advanced, we are non-idiots, and the whole point of this
exercise is we've realized that plurality is not a good system. We
oddly enough are not trying
to join the idiot team on this to win their admiration. Nevertheless
many people
have not yet realized that, and thus we'll face an obstacle to
overcome, hopefully
with education. But the fact is, polling data indicates, right now,
even with no education
beyond trying score system one in a sample election -- voters already,
right now, prefer
plain score over plurality.

Meanwhile, same polling data indicates, voters prefer plurality over
IRV, including
in Australian and in Britain and worldwide average.
Despite Richie's touting of the so-called advantages of IRV over score,
in particular re worry #2:
http://rangevoting.org/WhatVotersWant.html

So:
* Richie often does not know what a "problem" with a voting method is.
* Richie does not know which problems are common & which not.
He's often promulgated huge lies on that.
* Richie does not know which things voters want and which they do not.
He often announces he does, though.
* Richie often simply flat out lies.
Sorry to be a downer about Rob Richie, but that's how he's been for
many years. Decades. His only published article was incorrect and
indeed flat out
contradicted one of the most important theorems (the most?) in voting theory.
That was enough to ruin the career of any academic voting theorist,
but just his starting point.

Mark Frohnmayer

unread,
Oct 22, 2014, 3:06:57 PM10/22/14
to electio...@googlegroups.com
On Tue, Oct 21, 2014 at 10:48 PM, Warren D Smith <warre...@gmail.com> wrote:
> Yes William. The advantage of using scores is that the ranks can be
> trivially derived from the scores.

--except if equal...

Which is why it disincentives full naive exaggeration and bullet voting.
 

> This idea actually came out of a
> conversation with Rob Richie at the equal vote conference ... He suggested
> that approval plus top two could be done on one ballot (for overseas voters
> and the like) if there was an approval part and a rank order. By using
> score you get rid of the need for two separate ways of describing your
> vote.

--Oh great, it figures a bad idea like this would come from Rob Richie...

To be clear, Rob advocated a ballot that had approval and ranking. I have not won him over to solving both with a single score set.
 

>> As for the point that advocates of Score-off could be accused of "lying",
> maybe they could forestall that by saying "we believe plain Score is
> slightly better, but we want to compromise with another faction with whom
> we are not in full agreement, so as to achieve unity around a system we
> think will work much better than Plurality to deliver nearly equal
> political power to each citizen."
>
> Well partly it's for consensus,

--failed at that...

Working on it.
 

> but also I believe the system would
> discourage strategic score voting because the voter has to sacrifice his
> ability to express second round preference if he either fully naively
> exaggerated or fully bullet votes.

--groan.   "I know, let's intentionally make a system that reacts
badly to strategic voting,
because that'll discourage strategic voting!"     That plan never worked
in the history of the universe, far as I can tell...

You're oversimplifying. It doesn't react badly to strategic voting - it changes the nature of what a good strategic vote is. When the voter knows that two advance to the second round, all he needs to know is that one of his choices (preferably his actual favorite) makes it into the top two, and is then ranked higher than whoever else made it. To suggest that the algorithm used to count doesn't affect voter behavior and strategy flies in the face of all available evidence.
 

> Under what conditions that would beat out pure score is a question for the
> mighty oracles of election simulation (whether or not they think voters
> will actually vote in that less strategic way).

--arghh.  This all is a fix of a non-problem.  If it ain't broke, don't fix it.
To understand one reason it was a non-problem, consider
   http://rangevoting.org/AppCW.html
(very simple argument offered).

Look, you can keep arguing against it, or you can run the sim with a sane "strategy" and see what the numbers show.
This voting system is designed around the actually-genuine worry that the level of fracture in the reform community will ensure that no substantive reform passes ever. During the advocacy for the equal vote I have pitched literally thousands of real voters on the merits of various systems, and when I've explained score voting I've been directly confronted with the refrain "well, wouldn't you just score your one favorite maximum and give everyone else zero?" Not Rob Richie - some random woman in a coffee shop in SF. Whether on not this screws with the "perfect" algorithm or comes out great or whatever IS BESIDE THE POINT. Rather than explain how the voter is wrong, you can just say, "actually, since two make it to the final round, you are at a significant advantage if you support at least two candidates, and you can even express full support of your real favorite in the runoff between the two." Problem solved.
Yes, it's nice to have actual factual numbers to back up an argument either way.
 

But the fake second round idea, is just stupid.  It does not correct
for strategic distortions.  It does not correct for anything.  It in
fact takes valuable score data you'd painstakingly gathered, and
intentionally throws it in the garbage just at the crux
point of deciding the winner, intentionally using worse/less data for
that crux decision.

Again, you're missing the point. It corrects strategic distortions by making the ideal strategic vote more honest.
 

It also cavalierly sacrifices just about every logical property plain
score enjoys (simplicity, monotonicity, voting is better than not
voting, subdistrict consistency).

I thought you all were in favor of tossing logical properties in favor of Bayesian simulation.
 

So:
* Richie often does not know what a "problem" with a voting method is.
* Richie does not know which problems are common & which not.
He's often promulgated huge lies on that.
* Richie does not know which things voters want and which they do not.
He often announces he does, though.
* Richie often simply flat out lies.
Sorry to be a downer about Rob Richie, but that's how he's been for
many years. Decades.  His only published article was incorrect and
indeed flat out
contradicted one of the most important theorems (the most?) in voting theory.
That was enough to ruin the career of any academic voting theorist,
but just his starting point.

And... you're proving my point. One of the two of you has actually been successful in passing real reform in the real world. That it is a mathematically deficient solution is why finding a better consensus solution is important.

Warren D Smith

unread,
Oct 22, 2014, 7:55:18 PM10/22/14
to electio...@googlegroups.com
>> --groan. "I know, let's intentionally make a system that reacts
>> badly to strategic voting,
>> because that'll discourage strategic voting!" That plan never worked
>> in the history of the universe, far as I can tell...
>>
>
> You're oversimplifying. It doesn't react badly to strategic voting - it
> changes the nature of what a good strategic vote is. When the voter knows
> that two advance to the second round, all he needs to know is that one of
> his choices (preferably his actual favorite) makes it into the top two, and
> is then ranked higher than whoever else made it. To suggest that the
> algorithm used to count doesn't affect voter behavior and strategy flies in
> the face of all available evidence.

--huh? Actually, I suspect naive exaggeration strategy is highly
popular with voters in essentially ANY rank-order voting system,
REGARDLESS OF THE RULES, contrary to your final sentence here, and
know of no evidence contradicting that.

>> --arghh. This all is a fix of a non-problem. If it ain't broke, don't
>> fix it.
>> To understand one reason it was a non-problem, consider
>> http://rangevoting.org/AppCW.html
>> (very simple argument offered).
>>
>
> Look, you can keep arguing against it, or you can run the sim with a sane
> "strategy" and see what the numbers show.

--you can keep on ignoring all the arguments I find (did you even read
it?), and common sense itself... and to some extent, as I keep
pointing out but you keep ignoring, I already have simulated various
aspects of this, such as the fact that range voting is superior to
plurality voting even in 2-candidate elections, so therefore, your
"+instant" "improvement" is clearly a worsening, no further sim needed, QED.

>> But the fake second round idea, is just stupid. It does not correct
>> for strategic distortions. It does not correct for anything. It in
>> fact takes valuable score data you'd painstakingly gathered, and
>> intentionally throws it in the garbage just at the crux
>> point of deciding the winner, intentionally using worse/less data for
>> that crux decision.
>>
>
> Again, you're missing the point. It corrects strategic distortions by
> making the ideal strategic vote more honest.

--you do not know what the "ideal strategic vote" is.
In fact, nobody knows it. In fact game theory indicates there is no such thing
as optimal strategy in essentially all N-player games with N>2.

And if you do have some idea about strategy, the problem is, my sims
tend to have embedded in them, naive strategies. Not optimal
strategies, since they are not known and do not exist, in most cases.
So then if I ran a sim, you'd complain.

>> It also cavalierly sacrifices just about every logical property plain
>> score enjoys (simplicity, monotonicity, voting is better than not
>> voting, subdistrict consistency).
>>
>
> I thought you all were in favor of tossing logical properties in favor of
> Bayesian simulation.

--no, I'm not. Logical properties are a good guide. Especially for voting
system designers. BR simulations can be better, and to the extent we
can believe in the assumptions underlying the sim, they are indeed
more important since tell us
the thing we truly want to know. But they don't necessarily provide
you with much (sometimes not even any) understanding.

> And... you're proving my point. One of the two of you has actually been
> successful in passing real reform in the real world. That it is a
> mathematically deficient solution is why finding a better consensus
> solution is important.

--Richie has "succeeded" in making the world worse by passing
so-called reforms, which in most, maybe even all, cases have replaced
a better voting system -- plurality + genuine top-2 runoff -- with a
worse voting system: instant runoff.
He has accomplished this by lying to voters. In many cases, his
"reforms" were then repealed by disgusted voters after years of wasted
time, and/or phenomena that then occurred directly refuting
Richie+FairVote's lies, e.g Burlington VT, Aspen CO. In the case of
his biggest "success," San Francisco, polls show the voters want to
repeal IRV, having now realized it was a mistake, but they have not
yet been given the chance, so Richie's unwanted reform unfortunately
is staying, for the present, in SF, so the net result will be that SF
will have wasted 10, 20, who knows how many years.

Can you give me even a single case, where Richie has actually improved
any voting system in the real world? Not "changed to a worse system"
-- "improved."
There might be such a case. I would hope so. I just am not aware of one.

Now. Has range voting been prevented from being enacted, because of
voting system reformer "fracture"?
No. That is not its problem. You may think it is its problem, but if
so you are just wrong. Range voting has not been enacted, because it
has never been on a referendum ballot!

You tried to get something mildly like it on a ballot, and you failed.
Did you fail because
of fracture with Rob Richie? I don't think that was the cause, do you?
Polls indicate, if Score Voting is put on a ballot referendum, it will
probably pass.
At the first approximation, it is that simple.
It is just that there are considerable financial / labor obstacles to
getting it on a referendum ballot, that is all. Rob Richie conned and
lied his way to getting a lot of money, hence was able to get IRV on
some referenda, and sometimes it passed, unfortunately.

There also is indication, that voting systems should NOT be made complicated.
Here, yet another web page you can ignore, is some stuff about comprehension in
the real world of voting systems. You will observe that even a small amount
of complexity poses a problem in the real world with the substantial
fraction of voters who speak other languages, etc:

http://www.rangevoting.org/Comprehension.html

Mark Frohnmayer

unread,
Oct 22, 2014, 8:32:03 PM10/22/14
to electio...@googlegroups.com
On Wed, Oct 22, 2014 at 4:55 PM, Warren D Smith <warre...@gmail.com> wrote:

You tried to get something mildly like it on a ballot, and you failed.
Did you fail because
of fracture with Rob Richie?  I don't think that was the cause, do you?
Polls indicate, if Score Voting is put on a ballot referendum, it will
probably pass.

Indirectly. FairVote panned the proposal early on and regurgitated a bunch of their disagreement with CES. The Green Party in Oregon is composed of IRV disciples who then shat upon the initiative, and this gave funders the idea that there would be third party opposition (which was a killer of M65 in 2008) so they chose to support a measure that didn't explicitly mandate approval voting in the first stage because it had a ballot title that polled slightly better on naked read. Without funding support on such a short timeline it was not possible to gather the signatures, especially with a similar and funded measure in the field.

If you think that because the "idea" polls well it will pass at the ballot box, or even get on the ballot in the first place, you should try making it happen. The powers that be are HIGHLY resistant to change, and fracture within the election reform community is not helpful. Your continued screed below against Richie, justified or not, is just a further example of this.

At the first approximation, it is that simple.
It is just that there are considerable financial / labor obstacles to
getting it on a referendum ballot, that is all.  Rob Richie conned and
lied his way to getting a lot of money, hence was able to get IRV on
some referenda, and sometimes it passed, unfortunately.

There also is indication, that voting systems should NOT be made complicated.
Here, yet another web page you can ignore, is some stuff about comprehension in
the real world of voting systems.  You will observe that even a small amount
of complexity poses a problem in the real world with the substantial
fraction of voters who speak other languages, etc:

Sorry, but the assertion that this proposal is complex is absurd. It can be summed quite simply:

Rated IRV elects the majority favorite from the two highest rated candidates. (thanks Clay)

And its ballot looks EXACTLY the same as a score voting ballot. So if voters have high comprehension of a score ballot (as your link suggests), they'll likewise have high comprehension of a Rated IRV ballot.

Compare that to both a ballot and a description of IRV, or even top two runoff.
 

http://www.rangevoting.org/Comprehension.html

Mark Frohnmayer

unread,
Oct 22, 2014, 8:57:28 PM10/22/14
to electio...@googlegroups.com
To your other points:

On Wed, Oct 22, 2014 at 4:55 PM, Warren D Smith <warre...@gmail.com> wrote:
--huh?  Actually, I suspect naive exaggeration strategy is highly
popular with voters in essentially ANY rank-order voting system,
REGARDLESS OF THE RULES, contrary to your final sentence here, and
know of no evidence contradicting that.

You're just not giving any thought to the actual user experience of the system. If I'm a voter, and I know that differences in how I score candidates will be used to choose between them in the second stage, what is my motive to rate two candidates the same when I have a preference order between them?

There are very good actual reasons for the voting strategies people use in the real world, based on the counting algorithm. I vote for the most tolerable frontrunner in plurality voting because to fail to do so has shown over and over to have the potential to lead to the election of my least-favored candidate by way of the spoiler effect.

Likewise in IRV, that same spoiler effect is hidden, but still present, so over the long haul it actually makes sense for the voter to choose the more tolerable front runner as a first choice and shadow out any other lesser candidate he may prefer more.

These aren't naive strategies - they are actually the best strategy the voter has most of the time. Rated IRV does not suffer this defect.
 

--you can keep on ignoring all the arguments I find (did you even read
it?), and common sense itself... and to some extent, as I keep
pointing out but you keep ignoring, I already have simulated various
aspects of this, such as the fact that range voting is superior to
plurality voting even in 2-candidate elections, so therefore, your
"+instant" "improvement" is clearly a worsening, no further sim needed, QED.

Yeah, keep trotting that one out. I guarantee it will sound like nonsense to the lay voter, and that anyone who stands to lose power from the adoption of that system will absolutely skewer you in the eyes of the public.
 

>> But the fake second round idea, is just stupid.  It does not correct
>> for strategic distortions.  It does not correct for anything.  It in
>> fact takes valuable score data you'd painstakingly gathered, and
>> intentionally throws it in the garbage just at the crux
>> point of deciding the winner, intentionally using worse/less data for
>> that crux decision.
>>
>
> Again, you're missing the point. It corrects strategic distortions by
> making the ideal strategic vote more honest.

--you do not know what the "ideal strategic vote" is.
In fact, nobody knows it.  In fact game theory indicates there is no such thing
as optimal strategy in essentially all N-player games with N>2.

It should be possible to determine with simulation which strategies are better than others at reducing regret for voters using those strategies. The naive strategies are adopted in plurality and IRV in part because they work for those strategic voters.
 

And if you do have some idea about strategy, the problem is, my sims
tend to have embedded in them, naive strategies.  Not optimal
strategies, since they are not known and do not exist, in most cases.
 So then if I ran a sim, you'd complain.

Yeah, obviously I'd complain if you didn't at least try the simple strategy algorithm I suggested :-). If you did try the strategy I outlined and it showed that it performed badly, then I'd totally stop complaining, and if it worked well we could just argue about which strategy people would actually use.
 

>> It also cavalierly sacrifices just about every logical property plain
>> score enjoys (simplicity, monotonicity, voting is better than not
>> voting, subdistrict consistency).
>>
>
> I thought you all were in favor of tossing logical properties in favor of
> Bayesian simulation.

--no, I'm not.   Logical properties are a good guide.  Especially for voting
system designers. BR simulations can be better, and to the extent we
can believe in the assumptions underlying the sim, they are indeed
more important since tell us
the thing we truly want to know.  But they don't necessarily provide
you with much (sometimes not even any) understanding.

The problem with the logical properties is that they are binary -- either totally fulfilled or not at all. The nice thing about Rated IRV is that it balances competing criteria - later no harm and favorite betrayal, for example. And what the heck is subdistrict consistency anyway? It certainly sacrifices nothing in terms of simplicity for the voter. The ballot looks exactly the same as a score voting ballot. The algorithm for counting need not even be present on the ballot but could be explained during the public process as ballots are counted.

Clay Shentrup

unread,
Oct 22, 2014, 11:20:33 PM10/22/14
to electio...@googlegroups.com
On Tuesday, October 21, 2014 10:48:20 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
--arghh.  This all is a fix of a non-problem.  If it ain't broke, don't fix it.
To understand one reason it was a non-problem, consider
   http://rangevoting.org/AppCW.html
(very simple argument offered).

You may think it's very simple, but you have to understand that the vast majority of politically active people have zero understanding or really even awareness of social choice theory. Again, Steve Chessin (an MIT graduate and founder of Californians for Electoral Reform) argued against me at the Walnut Creek Democratic Club, that Score Voting basically untenable due to the fact that it could fail to elect a majority favorite.

It does not matter how insane that argument is, or how easily you can refute it. He *does not understand*. He failed to even understand Independence of Irrelevant Alternatives, and how it (combined with Condorcet cycles) can be used to prove that the Condorcet winner may not be the right winner.

Mark is trying to propose a system which is very good (certainly much better than IRV), and also passes the "dumb talking points" test. If you can say, "R-IRV always elects a majority winner", that helps satisfy the Dumb Talking Points test. It actually doesn't pass Chessin's test, since the majority favorite may be 3rd or worse in points, but at least you can say it "guarantees a majority winner". That one stupid phrase unfortunately seems to carry a lot of weight.

To "fix" a non-problem raised by a fool with zero evidence?

He may be a fool, but he has funding and talks a lot, as if he knows what he's talking about. So he has influence. If you refute his argument the "right" way, you win in the academic sense, but it doesn't matter, because no one understands that you won. Social choice theory is too counter-intuitive and esoteric for most people to understand it.

There are always idiots who make stupid criticisms of stuff.  The
correct response
is to explain why they are wrong.

Correct in the academic sense perhaps. Correct in terms of actually getting a better election system used? Different story.

* Richie often does not know what a "problem" with a voting method is.
* Richie does not know which problems are common & which not.
He's often promulgated huge lies on that.
* Richie does not know which things voters want and which they do not.
He often announces he does, though.
* Richie often simply flat out lies.
Sorry to be a downer about Rob Richie, but that's how he's been for
many years. Decades.  His only published article was incorrect and
indeed flat out
contradicted one of the most important theorems (the most?) in voting theory.
That was enough to ruin the career of any academic voting theorist,
but just his starting point.

I've been talking about how clueless and evil Richie is since the first week or two that I got into voting methods in 2006. But I've found it doesn't get me anywhere. There's no audience. No one realizes this is even an issue. They're worried about ebola. Or Renee Zellweger's plastic surgery. Thus "winning" an argument against Rob Richie is like winning an argument with a creationist you encounter on Twitter. No one cares one iota that you won the debate. It accomplishes nothing.

Clay Shentrup

unread,
Oct 22, 2014, 11:23:15 PM10/22/14
to electio...@googlegroups.com
On Wednesday, October 22, 2014 4:55:18 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
range voting is superior to plurality voting even in 2-candidate elections

In the eyes of every voter, you just lost the argument. It doesn't matter that you're actually right.

Warren D Smith

unread,
Oct 23, 2014, 12:16:07 AM10/23/14
to electio...@googlegroups.com
On 10/22/14, Mark Frohnmayer <mark.fr...@gmail.com> wrote:
> On Wed, Oct 22, 2014 at 4:55 PM, Warren D Smith <warre...@gmail.com>
> wrote:
>
>>
>> You tried to get something mildly like it on a ballot, and you failed.
>> Did you fail because
>> of fracture with Rob Richie? I don't think that was the cause, do you?
>> Polls indicate, if Score Voting is put on a ballot referendum, it will
>> probably pass.
>>
>
> Indirectly. FairVote panned the proposal early on and regurgitated a bunch
> of their disagreement with CES. The Green Party in Oregon is composed of
> IRV disciples who then shat upon the initiative, and this gave funders the
> idea that there would be third party opposition (which was a killer of M65
> in 2008) so they chose to support a measure that didn't explicitly mandate
> approval voting in the first stage because it had a ballot title that
> polled slightly better on naked read. Without funding support on such a
> short timeline it was not possible to gather the signatures, especially
> with a similar and funded measure in the field.

--sad. The US Green Party may know some stuff, but they are really
really stupid about
voting methods. Last I looked, they actually had IRV as a party
plank. How stupid is
that? Suicidally stupid. Why? Because IRV leads to 2-party domination.
If we adopt IRV, that will be suicide for the Greens. Just like
plurality leads to 2-party domination. (Meanwhile the Australian
Greens call for eliminating IRV...)

And by the way, computer sims are not as good as real life. Real life
shows IRV leads to 2-party domination. I don't see how to show that
in a computer sim without making assumptions that some would question.
In Australia, last 600 House elections with IRV, just 1 third-party
winner. In USA, last 10 thousand US House elections with plurality,
zero third party winners. (Independents can win in both USA &
Australia, though, usually former major party members.)

I contacted some top Greens about this. Mostly they did not want to
talk at all. Sometimes I could talk, and then mostly they said they
just were not interested in voting methods. Voting methods (they
usually said) are a minor issue, boring, of no consequence. Not like
the Important issues they are concerned with.

But, hello? All your important issues do not matter at all if you can
never win an election. It just does not penetrate their skulls. Same
thing happened with Libertarian Presidential Candidate Bob Barr. I
got to talk with him face to face about 2 hours. He gave me the same
spiel. Voting methods unimportant, not of interest. He refused to
put them in his campaign one iota. Because he was going to
concentrate on Important issues that mattered, because he was going to
win the US presidency. (He insisted he would win the presidency, and
I was too stupid to realize that.) And what was Bob Barr's top
issue? It was: politicization of the Justice Dept by G.W. Bush.

So eventually I came to have a theory. My theory is: in the USA, third party
high officials and candidates simply are not interested in winning.
In fact they
may actually fear and try to avoid winning. What they are interested
in, is maintaining some kind of delusion that they are important yet
sadly unappreciated. They need that feeling for their mental health
problems.

And in one case, admittedly a somewhat unusual third party, I got some
confirmation directly from its head. This was NY State's "working
families party." I was speaking with the two top guys in that party
in a room one time in Albany. They told me, right to my
face, they were NOT interested in running candidates. All they were
interested in, was in
attaching the working families "seal of approval" to one or the other
major party candidate, each election. They said if they did run a
candidate, it would take money & effort, plus they'd lose. That all
would be counterproductive as far as they were concerned. On the
other hand, they felt they could sort of use their "seal of approval"
thing to extort/bribe small concessions from NY politicians, thus
causing small amounts of positive change. "But," I said, "what if we
had a different voting system in which 3d parties could win? Wouldn't
that be better? Then running COULD be productive."
But they were not interested. In their view that was a far off,
uninteresting fantasy.
Sure, maybe it'd be nice in fairyland, but meanwhile they had
important things to do, and voting methods was NOT important, and NOT
worth them devoting any attention to.

In the Working Families' party's defense though, due to quirks in NY
state election system, they actually really are able to have an effect
(or at least have the delusion they have an effect) without ever
running a candidate. This option is not available
for most third parties in most US states, it is mainly only possible
in NY state.
Because NY allows them to get on ballot as a new party supporting an
old candidate already on ballot running as (say) a Democrat.

> If you think that because the "idea" polls well it will pass at the ballot
> box, or even get on the ballot in the first place, you should try making it
> happen. The powers that be are HIGHLY resistant to change, and fracture
> within the election reform community is not helpful. Your continued screed
> below against Richie, justified or not, is just a further example of this.

--look. I'm not going to pretend Richie is a great successful
well-informed reformer with wonderful original ideas.
He's had his chance for many years to make me think that.
Time after time he's shown over and over that he will not retract his
errors (lies) even after 20 years and massive evidence. I used to
document his published lies but stopped after I got to order 100.

Now, if you want to band with a lifelong liar and delusional guy, good
luck to you.

> At the first approximation, it is that simple.
>> It is just that there are considerable financial / labor obstacles to
>> getting it on a referendum ballot, that is all. Rob Richie conned and
>> lied his way to getting a lot of money, hence was able to get IRV on
>> some referenda, and sometimes it passed, unfortunately.
>>
>> There also is indication, that voting systems should NOT be made
>> complicated.
>> Here, yet another web page you can ignore, is some stuff about
>> comprehension in
>> the real world of voting systems. You will observe that even a small
>> amount
>> of complexity poses a problem in the real world with the substantial
>> fraction of voters who speak other languages, etc:
>>
>
> Sorry, but the assertion that this proposal is complex is absurd. It can be
> summed quite simply:
>
> Rated IRV elects the majority favorite from the two highest rated
> candidates. (thanks Clay)

--Sorry, that 1 sentence is not an adequate description. I wonder what
"highest rated" meant.
Anyhow, it is more complicated than score voting, that was all I was claiming,
and 10-15% of voters, right now, self-assessed understanding level,
after experience
of actual use, say they have either imperfect or very poor (basically no)
understanding of these voting methods right now:
approval, score, plurality, IRV.
How much depends on the method, ranging between 9.4% for score up to
around 14.4% for approval and IRV.

Do you think 10-15% of voters having that opinion, is a bad thing? I
do. The lesson
I draw from this is, even extremely small complexity changes affect
understanding a lot,
and we need very very simple methods.

Your arguments for why score+instant is "better" than score, well... it would be
hard to get people to say they "understand" them. Score has a lot of
use in the real world, movie ratings, olympic judging, school grades
0-100, many polls conducted for real-life elections, everybody has
experienced that ... while score-instant has had no use in the real
world yet, that I know. This makes it both pretty clear to me, and to
everybody, that score works ok, and they can instantly see it is
reasonable idea, and if they want to get reassurance of that they can
fall back on the score-style polls I keep trying to collect from round
the world to provide a databank of real experience, to back that up.

But why is score+instant something they instantly see is a reasonable
idea? Well, they can't see it. They have to get you to explain it to
them, about how it is propaganda move you designed to seek
"consensus", and they should just be happy
that wise minds are seeking consensus from dumber minds. Oh.

> And its ballot looks EXACTLY the same as a score voting ballot. So if
> voters have high comprehension of a score ballot (as your link suggests),
> they'll likewise have high comprehension of a Rated IRV ballot.

--The question was: "Le principe du vote par note vous semble-t-il clair?"
Not "is the ballot the same?"

> Compare that to both a ballot and a description of IRV, or even top two
> runoff.

--the poll studies found the French preferred score over plurality +
top two runoff,
but plurality is preferred over IRV.

Oho! I know, I guess that means we should try to incorporate elements
from IRV into score. Yeah, that'll make it more popular! Huh?
http://www.rangevoting.org/WhatVotersWant.html

--Incidentally, another interesting, albeit in the current environment
pretty unusual,
real-life election came to mind:
http://www.rangevoting.org/Romania2009.html

This election featured a "Condorcet cycle" so that no matter what
election method was used, you could then produce B and say "B is
preferred over your winner by a majority."
In fact, after you did that and declared B the new winner, I could
produce C and note he is preferred over B by majority, and so on
forever. (The whole idea of
"shifting to majority preferred winner" is self contradictory as this
real election illustrates.)
Anyhow,
* the score & approval voting winners both would probably have been Oprescu,
* the official winner was Basescu with Geoana 2nd (Oprescu officially
6th with only 3.2%
of vote)
* IRV winner unclear, one of {Geoana, Basescu, Antonescu}
* Borda: probably Geoana.

So in this case, score+instant presumably would have changed the
winner from Oprescu (score) to Geoana.

So this is of interest as a real life example of a flip. Who do you
think should
have won? Well, I'd go with Oprescu because I trust score & approval
voting, and since the pairwise-majority-preference in this case is
clearly complete
garbage because the condorcet cycle proves it contradicts itself. Why
throw away score voting (or approval) in favor of self-contradicting garbage?

Warren D Smith

unread,
Oct 23, 2014, 12:49:01 AM10/23/14
to electio...@googlegroups.com
Clay, I recognize that academic debates and winning referenda are not
necessarily the same thing, but the closest approximation we have to
the latter, is actual evidence from polls about what voting methods
voters prefer.

Over and over, self-appointed experts with no evidence, proclaim they
know from the Power Of Their MInds, what people want, while fools like
me who
actually have collected evidence, do not know, and whatever concerns I
have are merely "academic" (which again, they know from their Awesome
Internal Mental Power).

Therefore, they want to overturn 20 years of voting research to get nothing.
All because of the Awesome Power Of Their Self-Declared Internal
Confidence About What People Want.

And if Mark F. is concerned about fracture among voting reformers, I
point out, it is somewhat odd that almost his first move was to try to
create a new such fracture...

Clay Shentrup

unread,
Oct 23, 2014, 1:19:07 AM10/23/14
to electio...@googlegroups.com
On Wednesday, October 22, 2014 9:49:01 PM UTC-7, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:
Clay, I recognize that academic debates and winning referenda are not
necessarily the same thing, but the closest approximation we have to
the latter, is actual evidence from polls about what voting methods
voters prefer.

It doesn't really matter what voters prefer. It matters what the holders of political power prefer. That is, funded FairVote talking heads, and elected officials and some other activists. For instance, here in Berkeley, the endorsements of the Berkeley Democratic Club and the Sierra Club are two huge prizes, that have huge consequences for who wins elections. Those clubs are both sort of run by people who have an unusual amount of free time, and zeal for politics. In my experience, they are a crowd that is radically different from the general public.

Over and over, self-appointed experts with no evidence, proclaim they
know from the Power Of Their MInds, what people want

I'm talking about years of experience dealing with local pols and activists. I've heard their counter-arguments to Score Voting and Approval Voting so many times.

Your polls may tell us how a proposal might fair if it were to make it to the ballot (though even that's a stretch, since your polls don't reflect what happens when the voters have spent months on end looking at anti-Score-Voting posters on the train.) But they don't tell us about the viability of even making it to the ballot.

Therefore, they want to overturn 20 years of voting research to get nothing.
All because of the Awesome Power Of Their Self-Declared Internal
Confidence About What People Want.

I'm still quite passionate about Score Voting. I'm just expressing that R-IRV may actually fair better in the political/activist meme universe. And it's worth more investigation.

And if Mark F. is concerned about fracture among voting reformers, I
point out, it is somewhat odd that almost his first move was to try to
create a new such fracture...

I don't know that it's a fracture. After all, wouldn't you support R-IRV over IRV, TTR, or Plurality? Possibly even Approval Voting?

Warren D Smith

unread,
Oct 23, 2014, 1:50:12 AM10/23/14
to electio...@googlegroups.com
On 10/23/14, Clay Shentrup <cl...@electology.org> wrote:
> On Wednesday, October 22, 2014 9:49:01 PM UTC-7, Warren D. Smith (CRV
> cofounder, http://RangeVoting.org) wrote:
>>
>> Clay, I recognize that academic debates and winning referenda are not
>> necessarily the same thing, but the closest approximation we have to
>> the latter, is actual evidence from polls about what voting methods
>> voters prefer.
>>
>
> It doesn't really matter what voters prefer. It matters what the holders of
>
> political power prefer. That is, funded FairVote talking heads, and elected
>
> officials and some other activists. For instance, here in Berkeley, the
> endorsements of the Berkeley Democratic Club and the Sierra Club are two
> huge prizes, that have huge consequences for who wins elections. Those
> clubs are both sort of run by people who have an unusual amount of free
> time, and zeal for politics. In my experience, they are a crowd that is
> radically different from the general public.

--I doubt the "Berkeley Democratic Club" endorsement would matter much re
a Score Voting referendum on the CA ballot. Probably would matter a lot re
if I tried to run for office there.

> Over and over, self-appointed experts with no evidence, proclaim they
>> know from the Power Of Their MInds, what people want
>
>
> I'm talking about years of experience dealing with local pols and
> activists. I've heard their counter-arguments to Score Voting and Approval
> Voting so many times.
>
> Your polls may tell us how a proposal might fair if it were to make it to
> the ballot (though even that's a stretch, since your polls don't reflect
> what happens when the voters have spent months on end looking at
> anti-Score-Voting posters on the train.) But they don't tell us about the
> viability of even making it to the ballot.

--it will make it onto the ballot if you have money. It could be IRV,
approval, score, score+instant, doesn't matter too much, all will get
on the ballot if you have money.
All you need to do, in most referendum states, including CA, is gather
enough signatures
during/before certain dates. If you have enough money, you can
gather enough signatures to get essentially anything onto the ballot.
If it is more popular, that helps, but not a lot, re getting it onto
ballot.
A less popular one will require 2X as much money perhaps, that is all.
Whether the Berkeley Democratic Club endorses, is essentially
irrelevant to the question of whether you can collect enough
signatures, IF you have the money to pay the signature collectors.
Because the average Joe asked for a signature has no idea whether the
Berkeley Club endorsed it or not. Media also basically leaves all this
stuff uncovered during the signature collection stage. So that is all
it takes: Money.

Now once it gets on ballot, THEN whether it will win passage, has a
lot to do with
what the proposal is, and how it is worded, and what the media say about it.
The poll data I compiled is relevant to that.

Now if you do not have money, but are an incredibly
inspiring/charismatic organizer who can recruit a ton of free
volunteers to collect signatures, you might still be able to get it on
ballot in some states with low signature requirements. Far as I can
tell (and
I don;t know too much) examples of that are few. Mostly, it just
takes lots of money and grassroots campaigns to get things on ballot
simply do not work (or if they do, they really were only pretending to
be grassroots and really were not).

You can just pay commercial companies to gather signatures for you, they will
charge fixed amount of money per signature in many cases.

> I'm still quite passionate about Score Voting. I'm just expressing that
> R-IRV may actually fair better in the political/activist meme universe. And
> it's worth more investigation.

--fine.. investigate.

> I don't know that it's a fracture. After all, wouldn't you support R-IRV
> over IRV, TTR, or Plurality? Possibly even Approval Voting?

--I'm not sure. I think it ought to be better than Plurality, and probably IRV.
Not necessarily TTR or approval.

Mark Frohnmayer

unread,
Oct 23, 2014, 6:59:57 PM10/23/14
to electio...@googlegroups.com
> I'm still quite passionate about Score Voting. I'm just expressing that
> R-IRV may actually fair better in the political/activist meme universe. And
> it's worth more investigation.

--fine.. investigate.

> I don't know that it's a fracture. After all, wouldn't you support R-IRV
> over IRV, TTR, or Plurality? Possibly even Approval Voting?

--I'm not sure.  I think it ought to be better than Plurality, and probably IRV.
Not necessarily TTR or approval.

Well I started spelunking in the code and ...

now I can see why you're so resistant to trying a new strategy.

An alternate kludge to the concept would be to have two top score spots on the ballot -- the top score and the tie breaker.

0 1 2 3 4 5 5T

Not pretty, but it solves for your Max - 0.0000001 score.

Clay Shentrup

unread,
Oct 24, 2014, 1:54:16 AM10/24/14
to electio...@googlegroups.com
I have no idea why you're talking about the CA ballot. Even getting such an initiative on a CITY ballot would be a massive endeavor. In CA, most "home rule" cities (that can enact their own voting laws) are big enough for the signature hurdle to be significant.

Clay Shentrup

unread,
Oct 24, 2014, 1:54:50 AM10/24/14
to electio...@googlegroups.com
I'm going to buy Warren a copy of Refactoring for solstice.

Paul Cohen

unread,
Jan 7, 2015, 7:20:05 AM1/7/15
to electio...@googlegroups.com
From the standpoint of counting votes it does not matter what numbers are used for ranking  so long as they are contiguous.  For example, it is immaterial whether the voters are asked to choose a rank from the numbers 1,...,7 or from the numbers -3,...,+3.  I prefer the latter choice, however, since it makes it clear whether a vote is one of approval or disapproval.
 
So suppose that the voting population is mostly split between two parties - the Nuance Party and the Clarity Party.   Voters in the Nuance Party tend to see the world in shades of gray and to perceive all kinds of minor differences between candidates.   Post election analysis of the votes from the Nuance Party finds that on average they cast a vote of -2 for candidates they dislike but their votes are evenly distributed among -3, -2, and -1; their votes for candidates they like are similarly distributed pretty evenly among ranks 1, 2 and 3.

In contrast, the Clarity Party sees the world in black and white.  Candidates favored by the Clairity Party voters get a uniform +3 vote while those they dislike get a uniform -3 vote.  Somehow their candidates seem to win elections pretty consistently even though there are about as many Nuance voters as there are Clarity voters. 

This is why I prefer a system that allows only three ranks: -1, 0 and +1.   I have some additional thoughts but I'll put them in other comments. 

Paul Cohen

unread,
Jan 7, 2015, 7:32:41 AM1/7/15
to electio...@googlegroups.com
One problem that I see with the idea of using the first round of counting to reduce the field to two candidates is revealed by the question of how to handle a three-way tie for first place.   From nearly the opposite standpoint, why hold a runoff at all if there is a single clear winner in the first round?

If the same votes are re-tallied for the second round of counting, is it clear that the second round is apt to be any more decisive than the first round?  I would prefer turning to some entirely different counting scheme instead.  At the end of my article, What Might be the Best Voting System? I suggested a fairly simple way to do exactly this.

Paul Cohen

unread,
Jan 7, 2015, 7:41:53 AM1/7/15
to electio...@googlegroups.com


On Sunday, October 12, 2014 2:32:38 AM UTC-4, Clay Shentrup wrote:
...

There is also some sense in which it motivates a certain level of sincerity, given that your effect on the head-to-head matchup component is comparable in significance to your effect on picking the top two.
...

In the first round of voting, you might get lucky and find that both candidates in the runoff are candidates you favor; in that case you will probably be happy if you ranked those two candidates differently.  But it is just as likely that neither of the candidates in the runoff are candidates you favored and in that case you may regret not giving a maximum vote to the ones you did favor (since doing so would have helped them in the first round).

But more likely than either of these cases is that you approve of only one of the two candidates in the runoff count.  In that case you surely would regret it if you had not given the one you approve of a maximum vote and the one you disapprove of a minimum vote.

Clay Shentrup

unread,
Jan 7, 2015, 10:42:15 PM1/7/15
to electio...@googlegroups.com
Negative votes load to confusion and have other problems. There is lots of psychological research here to draw from. Be empirical.

Mark Frohnmayer

unread,
Jan 8, 2015, 2:51:28 PM1/8/15
to electio...@googlegroups.com
Clay, you should read the full link you just posted. It contradicts you regarding -1,0,1.

On Wed Jan 07 2015 at 7:42:15 PM Clay Shentrup <cl...@electology.org> wrote:
Negative votes load to confusion and have other problems. There is lots of psychological research here to draw from. Be empirical.

--

Paul Cohen

unread,
Jan 8, 2015, 8:32:22 PM1/8/15
to electio...@googlegroups.com


On Wednesday, January 7, 2015 at 10:42:15 PM UTC-5, Clay Shentrup wrote:
Negative votes load to confusion and have other problems. There is lots of psychological research here to draw from. Be empirical.


In the reference you provide, I find only two mentions of negative voting.  The first,

  1. Negative numbers such as in the range -100 to +100 would be a bad idea. That would lead to more voter confusion. (Imagine if some voter went for -100 instead of +100 by accident! Not good.) With partly-negative ranges, there is more chance for mis-entered and illegible data, and more likely to lead to voters wasting half their votes by only using the positive end of the scale, and harder on blind/disabled, and less familiar; and there seem to be no benefits compensating for these disadvantages. (Here's a strange psychological effect: Range voters like to award the score 0. It is the most popular score. If we make the range -10 to +10, however, then do range voters now like to award "-10"? No! Zero is still the most popular score.)

probably makes a good point.  The general public is not very comfortable with negative numbers and actually using them in a ballot might confuse some; there likely would be serious problems reading and interpreting ballots if they actually had to write + and - signs on the ballot. 

The other mention,

A contrary argument for {-1, 0, +1} score-set (Professor Brian W. Goldman)

1. Minimum possible number of options before it reverts back to approval voting, making it the easiest range voting system to implement using existing voting machines.

2. Values can be expressly mapped to words like "Reject", "Neutral", and "Approve" to get rid of the negative value problem, as voters would be hard pressed to mistakenly mark "Reject" where they mean "Approve."

3. Third parties get all the benefits of normal approval voting, and because of the "Neutral" value, third party candidates will likely be less "Reject"ed than the less liked major party. This is especially true in gerrymandered districts where a large number of voters are going to default vote for one major party, against the other, and likely be in the middle about third party candidates. As such this should still allow for the nursery effect.

I admit {-1, 0, +1} doesn't have the same discriminatory power of larger ranges, but it feels easier to implement, with less change from our existing voting system. It achieves many of the goals desired in Range Voting, even if it doesn't have quite the same power.

makes clear it is quite easy to avoid that conceptual problem as well as the penmanship issue.  

I don't have any actual psychological studies or even any epidemiological studies to quote but, like me, you have probably heard popular calls for having "none of the above" as an option on the ballot.  I would note that any balanced system (a system that allows an equal opportunity for casting a vote against a candidate as to vote for a candidate) actually does provide such an option.  Such a system is not at all confusing and the ballot can be quite simple.  

Brian Goldman

unread,
Jan 9, 2015, 10:40:30 AM1/9/15
to electio...@googlegroups.com

A contrary argument for {-1, 0, +1} score-set (Professor Brian W. Goldman)

If all goes according to plan I will be an assistant professor next fall, but for now I'm not. Warren, could you remove the title until then?

makes clear it is quite easy to avoid that conceptual problem as well as the penmanship issue.

Its important to note that once you switch to words translated into numbers there is no difference between {-1, 0, +1} and {0, 1, 2} except for how blank votes are treated. Which is a debate unto itself. Strategically the words and both number ranges are also identical. The only difference is honest / naive voters, so any decision should be based on balancing what psychology tells us is the best way to get accurate opinions from these types of voters with what can be done practically on voting machines. From what little I know, Likert scales are popular and they include expressing a negative opinion. 

Clay Shentrup

unread,
Jan 9, 2015, 1:58:34 PM1/9/15
to electio...@googlegroups.com
On Thursday, January 8, 2015 at 11:51:28 AM UTC-8, Mark Frohnmayer wrote:
Clay, you should read the full link you just posted. It contradicts you regarding -1,0,1.

I just re-read it and I don't agree. Perhaps you're referring to the statements by Brian W. Goldman? Those are all easily refuted. E.g. the nursery effect could be gotten with a 0-2 scale by making 1 the default if you leave it blank, instead of 0.

Mark Frohnmayer

unread,
Jan 9, 2015, 2:53:57 PM1/9/15
to electio...@googlegroups.com

Well that's all well and good. Just saying that link doesn't support your claim.


--

Clay Shentrup

unread,
Jan 9, 2015, 8:36:42 PM1/9/15
to electio...@googlegroups.com
On Friday, January 9, 2015 at 11:53:57 AM UTC-8, Mark Frohnmayer wrote:

Well that's all well and good. Just saying that link doesn't support your claim.

 Wrong. E.g.

"Negative numbers are not recommended."

Mark Frohnmayer

unread,
Jan 9, 2015, 10:03:42 PM1/9/15
to electio...@googlegroups.com

Contradicted later: "French voters in an exit poll study (see this) preferred both 3-point scales {-1, 0, +1} and {0, 1, 2} (about equally) over {0, 1} (approval voting)"

Clay Shentrup

unread,
Jan 10, 2015, 12:41:45 AM1/10/15
to electio...@googlegroups.com
On Friday, January 9, 2015 at 7:03:42 PM UTC-8, Mark Frohnmayer wrote:

Contradicted later: "French voters in an exit poll study (see this) preferred both 3-point scales {-1, 0, +1} and {0, 1, 2} (about equally) over {0, 1} (approval voting)"

That asks what laypersons preferred, not what actually worked better. They aren't election system experts. They aren't going to be aware, for instance, of the increased rate of error when using negative scores.

Mark Frohnmayer

unread,
Jan 10, 2015, 1:12:43 AM1/10/15
to electio...@googlegroups.com
Hmm. The claims in that page regarding the increased rate of error, etc, related mostly to arbitrary negative ranges - not -1, 0, 1. I'd be curious if the two books used as evidence (published 1954 and 1978) actually examine the specific Range3 case. But sure, handwave away what people actually say they like, because they're dumb laypeople and not experts. Or we could accept that the voters are the users and that their opinions on usability matter.

An interesting gem in there: (Here's a strange psychological effect: Range voters like to award the score 0. It is the most popular score. If we make the range -10 to +10, however, then do range voters now like to award "-10"? No! Zero is still the most popular score.) That would suggest that vote patterns would differ between -1,0,1 (dislike, neutral, like) and 0,1,2. 

Clay Shentrup

unread,
Jan 10, 2015, 9:18:16 PM1/10/15
to electio...@googlegroups.com
On Friday, January 9, 2015 at 10:12:43 PM UTC-8, Mark Frohnmayer wrote:
But sure, handwave away what people actually say they like, because they're dumb laypeople and not experts. Or we could accept that the voters are the users and that their opinions on usability matter.

The test for usability is whether people actually end up using a system correctly. This is empirical. It has nothing to do with opinion, which is the guess of a layman about what works better. 

Mark Frohnmayer

unread,
Jan 10, 2015, 9:38:55 PM1/10/15
to electio...@googlegroups.com

That's one test, sure. Whether they like it is another test, and whether they understand the mechanics is yet another.

Anyway you're giving zero evidence that range3 (-1,0,1) suffers from an "erroneous vote" problem of any kind. And hey, the Venetians used it for 500 years so it can't be all that bad. http://www.rangevoting.org/VenHist.html


--

William Waugh

unread,
Jan 18, 2015, 11:57:34 PM1/18/15
to electio...@googlegroups.com
Laying aside for the moment issues of selling it to people and issues of how long one would spend in the voting booth and how long the ballot is and the handling of write-ins, how well do you think the following scheme would work:

The ballot has two sections.  The top section lists the candidates and permits the voter to assign a score from 0 through 1 in increments of .1 to each candidate.  If the voter leaves a candidate unmarked, the system assumes the minimum score.

The bottom section lists every possible pair of candidates.  The voter is invited to indicate which candidate of the pair she would prefer to have win, if that pair of candidates make it to the instant runoff round of tallying, to be described below.

Tallying proceeds in two phases.

The top part of the ballots are treated as a score election and the totals added up for the candidates.

The top two in scoring proceed to the instant-runoff phase.

The tally examines the ballot entries that show the pair of candidates who made it to the instant runoff.  Whichever candidate is indicated as preferred on more of the ballots, where that pair is listed, wins.

Would this scheme produce a different winner than a simple Score election would in any case you can think of?

William Waugh

unread,
Jan 21, 2015, 12:51:17 PM1/21/15
to electio...@googlegroups.com
Here is another variant of instant score-off.  Would it accord the voters equal power over the outcome?  How would it change strategy relative to ordinary Score?  Would it behave pathologically to small changes, as IRV does?

Everyone fills a Score ballot.  Tallying proceeds in rounds.  For each round:

- Is there only one candidate?  If so, declare that candidate the winner and stop.
- If any ballots fail to mention a candidate, add to them a scoring of minimum for that candidate.
- Throw out any ballots that score all candidates equally.
- Linearly expand the remaining ballots so they take up the full range.
- Sum the scores of the candidates across the ballots.
- Eliminate the lowest-scoring candidate from the pool of candidates and from the ballots.
- Repeat.
Reply all
Reply to author
Forward
0 new messages