Jameson Quinn | 25 Aug 23:47 2014
Picon

Fwd: Empirical voting experiment: first numbers



2014-08-22 15:44 GMT-04:00 Kristofer Munsterhjelm <km_elmet <at> t-online.de>:
On 08/22/2014 01:05 AM, Jameson Quinn wrote:
Here I am surprised that Condorcet was considered more easy to understand than IRV. IRV advocates often say that the "remove the loser from the ballots and run again until someone gets a majority" is a very simple phrasing, and it certainly seems simpler than explaining Minmax. Did you explain the actual Minmax method or just Condorcet (the candidate that would beat every other candidate one-on-one wins)?

My explanation was just for Condorcet, though it did mention the word "minimax" in parentheses.


If you did explain Minmax itself, I am indeed surprised. I'm not going to complain, though! If the results are representative, that would be a serious counter to the "IRV is so easy" argument. The method itself is harder to understand according to your numbers, and if the advocates try to shift the goal to "as easy as 1-2-3", well, then Condorcet is just as easy because the front-end is the same.

I've actually contacted Fair Vote to get a new explanation for IRV. They've agreed to give me one, and I'll rerun a few sessions of the experiment with their wording, so that I can be accused of biasing the experiment with an intentionally poorly-written explanation.

...

However, that may also show that the Turkers aren't good at evaluating fairness. They consider Borda among the best, but we know about its extreme teaming incentive. OTOH, they also consider IRV in the Plurality class. I could understand either judgement, but both at the same time is quite unexpected.

Yes, you can certainly criticize their judgment here. Still, their perception is a fact we have to deal with.
 


I think these numbers are certainly interesting. To me, they clearly
bolster the case for joining forces behind approval activism, and for
eschewing IRV as an activist strategy; even for the majority of us who
see some other system as ultimately better than approval.

Right. Approval is a simple fix on Plurality, gives the best bang for the buck, and is easily understood. I think the greatest risk to Approval is a scenario where it is implemented, the chicken dilemma makes it dangerously unstable, and after having gone the wrong way a few times due to voters mis-anticipating each other, it is repealed in a similar way to how Burlington repealed IRV.

I'd agree that that's probably the biggest risk (besides "nobody pays any attention and it never happens"). How big is it? This experiment can help us see.

Note that this experiment is 100% in a chicken dilemma situation, with an unrealistically tiny number of voters, so insofar as the pathologies are avoided and/or accepted by the voters in the experiment, they'd be even less likely to be an issue in real life.


Maybe your strategy data will provide information on how realistic that scenario is.

It will certainly help us understand this question. I'll post about that as soon as I'm confident in my analysis.

Cheers,
Jameson

----
Election-Methods mailing list - see http://electorama.com/em for list info
Kristofer Munsterhjelm | 25 Aug 22:44 2014
Picon

Re: Empirical voting experiment: first numbers

On 08/22/2014 01:05 AM, Jameson Quinn wrote:
> As many of you know, I've been running an online voting experiment using
> human subjects from Amazon Mechanical Turk. I'm using a 3-candidate,
> 9-voter "chicken dilemma" scenario, with factions of 4, 2, and 3 voters:
> Cand:        X  Y  Z
> Faction:
> Red   4      3  1  0
> Green 2      0  3  2
> Blue  3      0  2  3
>      Size     payoffs
>
> Each group of 9 gets assigned factions and a voting method, and runs the
> election 3 times, with monetary payoffs proportional to the numbers
> above in the last two rounds. Then they answer a survey about how fair,
> easy to vote, and easy to understand they found the method, plus some
> demographic questions.
>
> The voting systems I have tested so far include approval, Borda,
> Condorcet (minimax), IRV, MAV (medians), plurality, and score. I plan to
> also test SODA very soon.
>
> My analysis of the outcomes and strategies is not yet ready to share
> here. However, I have some numbers on the survey results. I used a
> Kruskall-Wallis comparison test, appropriate for Likert-scale results
> like these. Here are the results for the question "How easy was it to
> *understand* {{methName}} (the voting system you used)?":
>
>
>          trt    means   M
> 1 approval  231.2388   a
> 2 borda     229.7286   a
> 3 score     205.4898  ab
> 4 MAV       195.5761 abc
> 5 plurality 172.9545  bc
> 6 condorcet 165.3897   c
> 7 IRV       118.8281   d
>
> The important thing about the above table are the letters at the end. If
> two systems share at least one letter in common, the differences between
> those systems are not statistically significant. So we can safely say,
> for instance, that Approval and Borda are easier to understand than
> Condorcet, but we can't tell whether MAV is as understandable as the
> former or as confusing as the latter.

Here I am surprised that Condorcet was considered more easy to
understand than IRV. IRV advocates often say that the "remove the loser
from the ballots and run again until someone gets a majority" is a very
simple phrasing, and it certainly seems simpler than explaining Minmax.
Did you explain the actual Minmax method or just Condorcet (the
candidate that would beat every other candidate one-on-one wins)?

If you did explain Minmax itself, I am indeed surprised. I'm not going
to complain, though! If the results are representative, that would be a
serious counter to the "IRV is so easy" argument. The method itself is
harder to understand according to your numbers, and if the advocates try 
to shift the goal to "as easy as 1-2-3", well, then Condorcet is just as 
easy because the front-end is the same.

> Now, one thing in this table gives me pause: the result for plurality.
> Sure, approval and Borda are simple and intuitive for most people; but
> are they really more so than plurality? I suspect that this may reflect
> a flaw in my experiment. People assigned to plurality may, as they take
> the survey, still be very hazy on what "voting method" means. If all
> they've ever seen is plurality, it's hard for them to imagine something
> different. So they may effectively be answering a different question...
> something like, "How easy was it to understand this experiment as a whole?"
>
> However, I think that the rest of the numbers here are reliable. So
> clearly, IRV is hard to understand, and Approval and Borda are easy.
>
> Now, for the question "How easy was it to figure out *how to vote* in
> {{methName}}?":
>
>          trt    means  M
> 1 approval  214.8881  a
> 2 score     211.1531  a
> 3 borda     206.8429 ab
> 4 MAV       195.7717 ab
> 5 plurality 190.6970 ab
> 6 condorcet 167.6250  b
> 7 IRV       163.4531  b
>
> Generally, rated methods are at the top, ranked ones are at the bottom;
> though Borda may be (perceived to be) an exception. Again, we can't
> entirely rely on the number for plurality.

Seems reasonable. I find ranking easier than rating (less to worry about 
whether I got the scale wrong), but I might well be in the minority.

> Finally, the question "How *fair* did {{methName}} seem to you?":
>
>          trt    means  M
> 1 borda     209.0000  a
> 2 MAV       206.8587  a
> 3 approval  206.0571  a
> 4 condorcet 200.2721  a
> 5 score     189.3776 ab
> 6 plurality 158.3939  b
> 7 IRV       157.6094  b
>
> Again, Approval comes in among the best, and IRV among the worst.
> Surprisingly, score is not significantly better than plurality/IRV
> (though it also isn't significantly worse than the best). In this case,
> though we still have to take the plurality numbers with a grain of salt,
> I think it's fair to give them some credence. Even if people were
> answering the question "How fair did the results of this experiment seem
> to you?", it's not unreasonable to lay whatever unfairness they saw at
> the feet of plurality.

However, that may also show that the Turkers aren't good at evaluating
fairness. They consider Borda among the best, but we know about its
extreme teaming incentive. OTOH, they also consider IRV in the Plurality 
class. I could understand either judgement, but both at the same time is 
quite unexpected.

> I think these numbers are certainly interesting. To me, they clearly
> bolster the case for joining forces behind approval activism, and for
> eschewing IRV as an activist strategy; even for the majority of us who
> see some other system as ultimately better than approval.

Right. Approval is a simple fix on Plurality, gives the best bang for 
the buck, and is easily understood. I think the greatest risk to
Approval is a scenario where it is implemented, the chicken dilemma
makes it dangerously unstable, and after having gone the wrong way a few 
times due to voters mis-anticipating each other, it is repealed in a 
similar way to how Burlington repealed IRV.

Maybe your strategy data will provide information on how realistic that
scenario is.
----
Election-Methods mailing list - see http://electorama.com/em for list info

Jameson Quinn | 22 Aug 01:05 2014
Picon

Empirical voting experiment: first numbers

As many of you know, I've been running an online voting experiment using human subjects from Amazon Mechanical Turk. I'm using a 3-candidate, 9-voter "chicken dilemma" scenario, with factions of 4, 2, and 3 voters:
           
Cand:        X  Y  Z
Faction:   
Red   4      3  1  0
Green 2      0  3  2
Blue  3      0  2  3
    Size     payoffs

Each group of 9 gets assigned factions and a voting method, and runs the election 3 times, with monetary payoffs proportional to the numbers above in the last two rounds. Then they answer a survey about how fair, easy to vote, and easy to understand they found the method, plus some demographic questions.

The voting systems I have tested so far include approval, Borda, Condorcet (minimax), IRV, MAV (medians), plurality, and score. I plan to also test SODA very soon.

My analysis of the outcomes and strategies is not yet ready to share here. However, I have some numbers on the survey results. I used a Kruskall-Wallis comparison test, appropriate for Likert-scale results like these. Here are the results for the question "How easy was it to understand {{methName}} (the voting system you used)?":


        trt    means   M
1 approval  231.2388   a
2 borda     229.7286   a
3 score     205.4898  ab
4 MAV       195.5761 abc
5 plurality 172.9545  bc
6 condorcet 165.3897   c
7 IRV       118.8281   d

The important thing about the above table are the letters at the end. If two systems share at least one letter in common, the differences between those systems are not statistically significant. So we can safely say, for instance, that Approval and Borda are easier to understand than Condorcet, but we can't tell whether MAV is as understandable as the former or as confusing as the latter.

Now, one thing in this table gives me pause: the result for plurality. Sure, approval and Borda are simple and intuitive for most people; but are they really more so than plurality? I suspect that this may reflect a flaw in my experiment. People assigned to plurality may, as they take the survey, still be very hazy on what "voting method" means. If all they've ever seen is plurality, it's hard for them to imagine something different. So they may effectively be answering a different question... something like, "How easy was it to understand this experiment as a whole?"

However, I think that the rest of the numbers here are reliable. So clearly, IRV is hard to understand, and Approval and Borda are easy.

Now, for the question "How easy was it to figure out how to vote in {{methName}}?":

        trt    means  M
1 approval  214.8881  a
2 score     211.1531  a
3 borda     206.8429 ab
4 MAV       195.7717 ab
5 plurality 190.6970 ab
6 condorcet 167.6250  b
7 IRV       163.4531  b

Generally, rated methods are at the top, ranked ones are at the bottom; though Borda may be (perceived to be) an exception. Again, we can't entirely rely on the number for plurality.

Finally, the question "How fair did {{methName}} seem to you?":

        trt    means  M
1 borda     209.0000  a
2 MAV       206.8587  a
3 approval  206.0571  a
4 condorcet 200.2721  a
5 score     189.3776 ab
6 plurality 158.3939  b
7 IRV       157.6094  b

Again, Approval comes in among the best, and IRV among the worst. Surprisingly, score is not significantly better than plurality/IRV (though it also isn't significantly worse than the best). In this case, though we still have to take the plurality numbers with a grain of salt, I think it's fair to give them some credence. Even if people were answering the question "How fair did the results of this experiment seem to you?", it's not unreasonable to lay whatever unfairness they saw at the feet of plurality.

I think these numbers are certainly interesting. To me, they clearly bolster the case for joining forces behind approval activism, and for eschewing IRV as an activist strategy; even for the majority of us who see some other system as ultimately better than approval.

I'll be sharing more numbers from this experiment as I have them ready. Also, if anybody here wants access to my raw data, I'd be happy to share; though of course, I'd want you to duly cite me if you use them for anything.

Cheers,
Jameson
----
Election-Methods mailing list - see http://electorama.com/em for list info
Jameson Quinn | 10 Aug 17:33 2014
Picon

New paper: "~25% intrinsic-honest voters"

Kawai and Watanabe 2013 uses what looks to me to be pretty reasonable statistical methods and assumptions to estimate that voters for the Japanese House of Representatives (plurality single-member districts) are between 68% and 83% intrinsically strategic. Crucially, their methods can estimate the number of intrinsically "strategic" voters who end up voting honestly because that was the best strategy for them; it turns out that they estimate that 92-98% of intrinsically "strategic" voters in this sense actually vote honestly (that is, 95-98% of all voters are voting honestly in their data).

In order for their model to work, they need to be comparing the results of similar sub-regions (like precincts; actually, they use small municipalities) in different electoral districts. They must also assume that different candidates from the same party are essentially similar. There must also be sufficient votes for more than 2 parties for their model to work with. I doubt that they could have gotten a worthwhile estimate if they'd used US data.

I haven't read the paper carefully enough to be 100% sure, but based on a quick skim, it appears that they're assuming that the proportion of strategic voters is constant across ideological groups. That's probably necessary in order for their parameters to be identifiable/estimable for the data they have, and close enough to true for their results to be valid. However, this does mean that their work doesn't give any evidence one way or the other about how or whether strategic proportion varies across ideology.

Still, these numbers seem pretty reasonable to me. It's certainly useful to have an "empirical" number that I can plug into my VSE (aka BR) simulations. In particular, the most realistic scenarios are in the range from 75% strategic/25% honest, 50% strategic/50% one-sided strategic. That's a considerably narrower range than if you have to consider any combination of strategy, honesty, and one-sidedness. Someday soon I'll re-run my VSE sims focusing on these numbers and report what I find here. 

If anybody wants to read the PDF of this paper but doesn't have access, email me privately, and I'll send you a copy.


----
Election-Methods mailing list - see http://electorama.com/em for list info
robert bristow-johnson | 24 Jul 20:05 2014

Re: Voter strategising ability


Jameson, i am assuming you meant this for the list.

*wow*!  i would have never expected a real-world election where it would 
have mattered (in the outcome) whether it was a Shulze or Ranked-Pairs 
Condorcet method.  not that the Romania 2009 was *any* Condorcet, but 
was it STV and is the ballot data available?  otherwise i would ask, how 
do we know how be the Smith set was?

bestest,

r b-j

On 7/24/14 12:31 PM, Jameson Quinn wrote:
> I believe that cycles in real-life, contentious Condorcet elections 
> would be rare — on the order of 1-2% of elections, or a bit lower if 
> your data include "elections" with only one serious candidate. 
> However, when a Condorcet cycle does happen, it basically implies the 
> existence of at least three separate, more-or-less coherent factions 
> in the electorate. In that case, there's no particular reason that 
> there shouldn't be 2 "quasi-clone" candidates from one of those 
> factions. Thus I'd expect 4-member Smith sets to be almost a third of 
> all Smith sets larger than 1; not "*never*" as Robert suggests. In 
> fact, it is possible that Romania 2009 
> <http://rangevoting.org/Romania2009.html> had a 4-member Smith set.
>
>
> 2014-07-24 0:11 GMT-04:00 robert bristow-johnson 
> <rbj <at> audioimagination.com <mailto:rbj <at> audioimagination.com>>:
>
>     On 7/23/14 2:17 AM, Juho Laatu wrote:
>
>         On 20 Jul 2014, at 22:48, Kristofer
>         Munsterhjelm<km_elmet <at> t-online.de
>         <mailto:km_elmet <at> t-online.de>>  wrote:
>
>             Discussion about which kind of strategy is most likely to
>             happen can go on forever without data. Even if there is
>             data, it is quite easy and/or tempting to explain it away
>             as not being representative of what would happen under an
>             ordinary election. As long as that's possible, it's really
>             hard to convince someone who is worried about burial not
>             to be, or vice versa.
>
>         Unfortunately we don't have data from very many Condorcet
>         elections.
>
>
>     but the *data* doesn't give a rat's ass *how* it's counted or
>     tabulated.  can't we use the data from all ranked-choice elections
>     (which, in government, would be IRV or RCV or AV or STV or Hare)
>     and see how they would work out with Condorcet-compliant rules?
>      like we did for Burlington 2009.  that was a 4-way election close
>     enough that the Plurality (of 1st choice votes) winner, the IRV
>     winner, and the Condorcet winner were three different candidates.
>      and yet there was *no* cycle.  not even close to a cycle.
>
>     are there any other ranked-choice elections where media or
>     research could access the anonymous ballot data and see if there
>     would have been a cycle and then see how Shulze and Tideman and
>     Minimax and Kemeny would have been different?  i think we would
>     virtually never see a cycle.  and *more* than virtually, i think
>     we would *never* see a cycle with more than 3 in the Smith set.
>
>
>           And those elections have been quite non-competitive. So we
>         don't know very well what would happen (in different
>         societies) in competitive Condorcet elections.
>
>
>     but with ballot data in public records (and a little bit of
>     computer programming), we *should* be able to use all that IRV
>     ballot data and see what might happen in hypothetical
>     Condorcet-compliant elections.
>
>
>     -- 
>
>     r b-j rbj <at> audioimagination.com <mailto:rbj <at> audioimagination.com>
>
>     "Imagination is more important than knowledge."
>
>
>
>     ----
>     Election-Methods mailing list - see http://electorama.com/em for
>     list info
>
>

--

-- 

r b-j                  rbj <at> audioimagination.com

"Imagination is more important than knowledge."

----
Election-Methods mailing list - see http://electorama.com/em for list info
Jameson Quinn | 18 Jul 00:40 2014
Picon

Request: Voting method blurbs for research

I am continuing my research on voting methods on Amazon Mechanical Turk. I'm currently looking at 8 voting methods:
approval, Borda, "Condorcet" (minimax), IRV, MAV (Majority Approval Voting; that is, Bucklin/Medians with ABCDF grades, breaking ties by above-median votes), plurality, score (0-10), and SODA.

The full protocol is outlined below, but one aspect is that, at the end of the experiment, each subject takes a quick survey, including their feelings about the voting system they used. I'd like to test out different system descriptions, to see if how the description impacts those feelings. So, I'd like blurbs to describe how you'd vote, and how votes are counted, in each of these systems. Of course, clarity and brevity are desirable.

So, I have my own blurbs already, but anybody else who can write alternate blurbs for me, I'd appreciate it. You don't have to cover all 8 systems if you don't want, but please do as many as you can; at least 5. You can use html trickery (collapsible sections, popups, tables, images), or not; however you want.

Here's how the experiment works, as described on the landing page:

Please do not take this HIT until you are told to. Press "Next" below when you're ready to.

Voting Experiment

This is an experiment on voting. 18 voters like you, divided into three groups, will decide between three options. Depending on the winning option, participants will earn extra pay of up to $2.40 (paid within hours as a "bonus" in AMT). The average pay for each participant will be at least $2.67, and depending on your luck and skill you may earn up to $4.00 in total. We will be running this experiment several times, using different voting systems, but you may only participate once.

In order to ensure enough simultaneous participants, we will be starting this experiment at a defined time. Until the countdown finishes, you can only view steps 0 (this screen) and 1 (consent form). Press "next" below to see the consent form and countdown. If you leave this window open, when the countdown completes, a sound will play ("voting experiment starting") and a "consent and join" button will appear. At that point, we will accept only 18 subjects per experiment run, on a first-come first-served basis. We ask that you only "accept" the HIT with Amazon after you are allowed into the experiment. (But we have made more than 18 HITs available for idiot-proofing.) Only the 18 workers allowed in will be paid for each run of the experiment.

Process

Step Name Time Payout Explanation

0

Overview

0-0.5 mins See an outline of the experiment (this stage).

1

Consent

0-2.5 mins Understand your rights, wait for the experiment to begin, and informed consent.

2

Scenario

1-15 mins Understand how much you and other voters will earn depending on which of the virtual candidates wins.Also, wait until the experiment fills up before proceeding (a sound will play when ready).

3

Election method practice

1-1.5 mins Learn and practice the election method to be used.

4

Practice results

1-0 mins See results of the practice election: the winner and how much you would have been paid.

5

Voting round 1

0.5-1 mins Vote. You will be paid based on results.

6

Payout round 1

0.5-0.5 mins $0-$1.20 See results of the round 1 election: the winner and how much you will be paid. (Payments will arrive within 1 day)

7

Voting round 2

0.5-1 mins Vote. You will be paid again based on results.

8

Payout round 2

0.5-0.5 mins $0-$1.20 See results of the round 2 election: the winner and how much you will be paid. (Payments will arrive within 1 day)

9

Survey

2-5 mins $1.60 4-5 simple questions each about:
  • you (gender, country, etc)
  • the voting system you used (on a 0-7 scale)
  • your general comments about the experiment

10

Debrief

0-0.5 mins Thanks for participating, and a simple explanation of what we hope to learn from this study. Submit job and receive base pay.

Total

7-16 mins $1.60-$4.00 The total time will mostly depend on how quick the other turkers in the experiment are. (15 minutes are allowed on step 2 for the experiment to fill, but that is usually much quicker.)

Press the button below to see the consent form and wait for the experiment to start.



----
Election-Methods mailing list - see http://electorama.com/em for list info
Michael Ossipoff | 14 Jul 15:49 2014
Picon

Retiring from voting systems, quitting EM

I only mention that because I want it to be clear that, when I don't reply to a posting, I'm not being rude--it's just that I haven't received the posting, because I'm no longer subscribed to EM, and have retired from voting-systems.
 
Of course there's  one temporary exception to my voting-system retirement: If there are one or more replies to the messages that posted to EM a few minutes ago, and those replies are posted within a few days, then I'll reply to them (if they're civil). But, when such discussion has concluded, or if 48 hours pass without replies to the messages I posted today, then I'll unsubsubscribe EM, as part of my complete and final retiredmnt from voting-systems.
 
Michael Ossipoff
 
----
Election-Methods mailing list - see http://electorama.com/em for list info
Gervase Lam | 12 Jul 21:34 2014

Voter strategising ability

I'm a bit concerned about the possible strategic manoeuvres that each
faction of ballots performs in the description below.

I can't see each faction unilaterally carrying such manoeuvres out.
Surely it would require thinking that is beyond the lay voter?

Alternatively, I think a candidate steering voters into the appropriate
strategic voting (as suggested in the past on this list) can be risky.
Other candidates and even the "neutral" media (e.g. editorial
reporters/commentators) would criticise the fact a that this is an abuse
of the democratic vote process (i.e. voters should vote in an entirely
sincere fashion).

Assuming that people are happy with strategic voting, then data would
need to be provided for strategising.  I expect this would be in the
form of polls/voter intention surveys being carried out market research
companies.

The voter intention surveys for the UK European 2014 elections that I
saw were more than just people answering question "who would you vote
for".  They included asking questions like how did you vote in previous
elections, age and even whether they were private or public sector
employed.

With the UK General election in mind for May 2015, the surveys included
questions that were basically on the lines of "would you vote
differently if it were the General Election tomorrow instead of the
European election and if so who would you vote for" [FYI, the answers
were generally yes to this].

I assume the "normal" person (whoever that is) would ignore the 2015
election 'poll' if they were voting in the European 2014 election.  [Why
did the surveys include the 2015 election when the one closest in time
is the 2014 election?  Well, it's because to the UK voters, the 2015 is
by far the more important.]

Also interestingly, as well as asking those surveyed who they would vote
for, most of the surveys also asked how likely they would vote in each
of the 2014 and 2015 elections on a scale from 0 to 10.  With the data
provided in most of those surveys, I would find it easy to use a
spreadsheet and use the scale to weight each vote.  Some of the surveys
did that while others only considered those who would definitely vote in
their count.

In the UK, the European count is done with multi-member districts.
However, of the four or more surveys I looked at, only one reported
their polling against the correct districts.  For the other surveys, I
only found national counts or counts for regions that only vaguely
matched the geographical districts!

Given the above, I really find it hard to see a good proportion of
voters doing the correct strategic calculations.  Also, this is only for
plurality voting!  I don't really know if voters are going to handle
rank voting strategising as per the below.

May be they could download some sort of spreadsheet from online that
could do the strategising?  But I don't think it would be flexible
enough to handle any scenario.  Also, would the voters bother checking
how the spreadsheet worked just in case a rogue spreadsheet was
downloaded?

I suppose to conclude all this, I'm just wondering if a voting method
should handle the situation where voters carry out bad strategy!!?  I
can't see any voting method being able to handle that!

A variation to the above question is, can there be a voting method that
can handle voting where each faction carries out their own strategic
voting AND can handle voting where within each faction voters carry out
their own strategic voting.  The latter may significantly be due to the
fact the voters are floating voters who don't tow the candidate/party
line.  Therefore, they would have different opinions about the other
candidates.

In this current climate, I think most voters would vote sincerely with
practically any 'reasonable' voting method.  But I don't know about the
future.  I think the voting method should be solid for the future as it
can be extremely difficult to change a voting method once one is
instituted.

Ideally, I think voters should always be voting sincerely as
strategising (both for good and badly executed strategies) makes things
complicated.  But Gibbard-Satterthwaite shows this can't happen.

Just my stream of consciousness thoughts on all this...

Thanks,
Gervase.

> [EM] Concerning Chicken Proof Smith compliant methods
> Forest Simmons fsimmons at pcc.edu 
> Fri May 9 17:48:53 PDT 2014 

        ________________________________________________________________
> Suppose that max(y, z) < x < y+z,  and that a sincere summary of the voter
> preferences is
> 
> x: A>C
> y: B>C
> z: C>A
> 
> These sincere preferences could not constitute an informed ballot profile.
> Why not?  Because it would not constitute a strategic equilibrium:  The A
> faction could unilaterally truncate C, and thereby win the election.
> 
> How do we know this without knowing what election method i being used?
> Well, we are assuming that the metho is chicken proof, an if so, candidate
> A would be elected wih the following ballot set:
> 
> x: A
> y: B>C
> z: C
> 
> And untruncating A in the C faction could not make A lose in any of the
> methods we have been considering, even the non-mono-raise ones like Benham
> and Woodall.
> 
> x: A
> y: B>C
> z: C>A
> 
> But this position is not a strategic equilibrium either, since th B action
> could benefit y unilaterally raising C to equal top:
> 
> x: A
> y: B=C
> z: C>A
> 
> in which case C would be the winner.
> 
> What's more, this position is a strategic equilibrium, as is the posiiion
> 
> x:A>C
> y:B=C
> z:C>A
> 
> which is just one move from the sincere preferences, and hence the most
> likely equilibrium position.  Under pefect information it is the strongest
> game theoretic solution.
> 
> In summary, if sincere preferences are
> 
> x: A>C
> y: B>C
> z: C>A,
> 
> then rational ballots will be
> 
> x: A>C
> y: B=C
> z: C>B
> 
> So the sincere Condorcet preference is also the strategic ballot CW.
> 
> 
> In general (at least in the case of three candidates) if candidate X is the
> sincere Condorcet preference, candidate X will also be the ballot CW for
> ballot voted by rational voters under complete infomation.
> 
> In particular, the ballot set
> 
> x: A>B
> y: B>C
> z: C>A
> 
> will never be voted by rational voters when there is a sincere Condorcet
> preference.  Nor will
> 
> x: A
> y: B>C
> z: C,
> 
> Why not?  Because they are not strategic equilibria, except possibly in the
> absence of any true Condorcet preference.
> 
> So why do we pay so much attention to these non-equilibrium ballot sets?
> Precisely because we want to make sure that they are not equilibrium
> positions potentially rewarding arm twisting strategy, like the chicken
> strategy.
> 
> Forest

----
Election-Methods mailing list - see http://electorama.com/em for list info

Ross Hyman | 12 Jul 14:07 2014
Picon
Picon

Re: Random Ballot Condorcet

Sadly, the random ballot Condorcet method I posted about earlier does not in general elect the highest ranked Smith candidate on the random ballot when there are more than three Smith candidates.

The following method will work though.
 Chose a random ballot. If it is not complete, draw others to break ties until there is a complete ranking.  Elect the highest ranked candidate for which there is a beat path from it to every other candidate.

This can be formulated in a way similar to the previous method:  Candidates are either hopeful or discarded.  All candidates are initially hopeful. All candidates, hopeful and discarded are available to be used in beat paths.  Consider the two lowest ranked hopeful candidates.  Discard the lower ranked of the two if there is a beat path from the higher ranked candidate to the lower ranked candidate. (And it doesn't matter what its strength is or if there is a stronger beat path going the other way.)  If there is no beat path from the higher to the lower candidate and there is at least one beat path from the lower to the higher candidate then discard the higher candidate.  If there is no beat path either way, then define a beat path from the higher to the lower candidate and discard the lower candidate. Repeat until one hopeful candidate remains.  Elect that candidate.

The N seat Random Ballot Condorcet STV method is similar.  Construct a ranking of every relevant set of N candidates from random ballots (you will generally need more than one ballot even if all candidates are ranked. I will give a mechanism for doing this in another post.)  Elect the highest ranked candidate set for which there is a beat path from it to every other candidate set.

This method differs from a fully deterministic method because the only elections that must be considered to create the initial set of beat paths are all elections with N+1 candidates for N seats.  For each of these elections, create beat paths from the winning N seat candidate set of that election to each of looser sets of that election that is the winner of at least one N+1 candidate set election. Other beat paths are created as needed as the method proceeds when the lowest ranked hopeful candidate sets have no beat paths between them.    















On Tuesday, May 20, 2014 5:55 PM, Ross Hyman <rahyman <at> sbcglobal.net> wrote:


A better random ballot Condorcet method is: Chose a random ballot (and if it is not complete, draw others to break ties until there is a complete ranking).  Eliminate the pair-wise loser of the two lowest ranked candidates.  Repeat until one candidate remains.  Elect that candidate.

I believe it has the following desired properties: monotonic, clone independent, only Smith candidates get a non-zero probability of being elected, independence of zero probability alternatives, and it requires the fewest number of pair comparisons and chooses the candidate that tends to be higher ranked than the previous version. In the three candidate case, if there is a cycle, it will always choose the top ranked candidate from the random ballot.  

One can form a complete social ranking by starting from the lowest ranked candidate and moving candidates down if they lose to the one below it.  The social ranking from the previous method is equivalent to starting from the highest ranked candidate and moving candidates up if they beat the one above it.



   

 







On Wednesday, May 7, 2014 6:51 PM, Ross Hyman <rahyman <at> sbcglobal.net> wrote:


Random Ballot Condorcet:  Choose a random ballot.  Elect the lowest ranked candidate that pairwise beats all higher ranked candidates.

Has this method been discussed before?  I believe that the following are true:  It will always elect a Condorcet candidate if there is one.  Otherwise it will elect a member of the Smith set with some nonzero probability for each member of the Smith set.  Non-Smith set candidates will have zero probability of being elected.  It is monotonic in that raising a candidate on some ballots cannot decrease its probability of being elected.  It is clone proof in that the probability of electing from the clone set is independent of the number of clones in the set. It is independent of irrelevant alternatives in that deleting a candidate with zero probability of winning cannot effect the probabilities for electing other candidates.  
 

 







----
Election-Methods mailing list - see http://electorama.com/em for list info
Peter Zbornik | 1 Jul 19:36 2014
Picon

Software for STV elections - STVBallot

Dear all,

in the Czech Greens we developed this open source vote-entry and
counting program as an App in Chrome called STVBallot.
It allows for vote counting in a network without any dependance on the
internet and without client-server architecture.
Unfortunately the program is in Czech.
The vote-counting algorithm is standard STV, with the possibility to
combine ranked election of some candidates and the incorporation of
gender quotas, without resorting to "guarding" some candidates
(candidates are quoted after those with higher support are elected).

Download STVBallot into Chrome here:
https://chrome.google.com/webstore/detail/stv-ballot/nckikmpbpcjagehockhckflmmflkifbo

STVBallot was programmed by Vaclav Novak (in copy).
I took part in the analysis, testing and the integration of the
program in our party.

Best regards
Peter Zbornik
----
Election-Methods mailing list - see http://electorama.com/em for list info

⸘Ŭalabio‽ | 28 Jun 23:40 2014

¡Mathematics!] ¡TauDay! [/¡Mathematics!]

	⸘Howdy‽

	We are all into ElectionMathematics, so I figured that you all might like this post about TauDay I wrote:

		http://TauDay.Com

	¡Happy TauDay!

	Today is TauDay.  τ (Tau) is the 1 true CircleConstant.  It is TauDay because it is the 28th of June and τ
(Tau) is:

	c/r 	 = 	 τ 	 ≈ 	 6.28318530717959

	In ISO-8601, the 28th of June is:

	YYYY-06-28

	Because the format is:

	YYYY-MM-DDThh:mm:ss

	The first 3 significant figures of τ (Tau) are:

	6.28

	Which looks like:

	YYYY-06-28

	Which is why the 28th of June is TauDay.

	For TauDay, I converted τ (Tau) in to Balanced Ternary:

		1T0 . 10T,T0T,110 , 0T1,10T,T0T , 1TT,000,001

	http://WikiPedia.Org/wiki/balanced_ternary

	http://TauDay.Com

 
	¡Peace!

--

-- 

	“⸘Ŭalabio‽” <Walabio <at> MacOSX.Com>

Skype:
	Walabio

An IntactWiki:
	http://intactwiki.org

	“You are entitled to your own opinion, but you are not entitled to your own facts.”
	——
	Senator Daniel Patrick Moynihan

----
Election-Methods mailing list - see http://electorama.com/em for list info

Gmane