Talk:Electoral system/Archive 6

Rearrange table with Favorite Betrayal as first column
The Favorite Betrayal Criterion, more than any other, directly indicates whether an election system provides perverse tactical incentives to vote dishonestly. Ranking another candidate ahead of your honest favorite is perhaps the surest way to guarantee that the final tallies do not reflect the honest opinions of voters. We should make it easy for people to see which election systems encourage doing so, and which do not. As such, I propose rearranging the table of criteria to put the favorite betrayal criterion in the first column. What do you think? —Qaanol (talk) 23:40, 29 December 2012 (UTC)


 * All right, with no objections, I have gone ahead and made the change. I also made the entry for Range Voting in the Majority Criterion column match the entry for Approval Voting in that same column, because those two election systems behavior the same with regard to that criterion. Namely, a majority bloc of voters with the same favorite candidate, can guarantee that candidate wins if they know they are part of such a majority (as, for instance, pre-election polls could tell them.) And I fixed a different cell that had no text and an invisible border. —Qaanol (talk) 04:20, 1 January 2013 (UTC)


 * Um.
 * These edits are clearly good-faith, and probably point the way towards improvements. But as they stand, I oppose them, and I'd ask you to revert so we can take this a little slower. One at a time:
 * 1. FBC at front
 * Personally, I agree that FBC is one of the most-important criteria, and I'm happy to have it at the front. However, it scarcely even exists in the literature, and the FBC article has been deleted more than once. So I think your move is controversial. Moreover, it breaks the division of the criteria into categories, as shown in the previous section explaining the criteria meanings. So for both of those reasons, I appreciate your WP:BOLD action, but I think it needs further discussion.
 * 2. Range/Score voting and the majority criterion: While you are correct that Range meets the Majority criterion if the majority votes strategically, just as approval does, that is not what "rated" means here. The "rated" majority criterion simply states: "If a candidate is the only one to recieve the highest rating/ranking on a majority of ballots, that candidate wins". This criteria is not met by range: 2 ballots that vote X100 Y99 can be overcome by one ballot that votes X0 Y100, and X will lose. (Yes, the foregoing votes are not strategic. However, that issue is covered in the "strategic, majority Condorcet" column, which Range already ranked as passing.)
 * I will not be around for the next week or so, but I'm reverting your changes. Please discuss this further. If you want someone to argue with while I'm gone, you might have luck by pinging MarkusSchulze or anybody else you see participating above on this talk page. Homunq (࿓) 12:07, 1 January 2013 (UTC)


 * Thanks for the info, I did not realize the FBC was rare in literature. I would have expected it to be well documented and supported, because it seems one of the most important criteria. After all, if an election system encourages insincere votes, then it does not matter how fancy the tallying method is, because garbage in, garbage out. In any case, I'll leave your version in until we get further discussion. —Qaanol (talk) 05:59, 2 January 2013 (UTC)


 * You should read the literature on what happens with sincere and insincere voting in practice. For example, IRV violates favorite betrayal, but I don't know of any real-life examples of it affecting how people vote despite so many IRV elections to example. On the other hand, violation of later-no-harm absolutely leads to truncation of rankings (Bucklin) or approvals (approval voitng) in the handful of places that have tried system that violate it... My theory is that a lot of voting methods folks are better at understanding math than human psychology.RRichie (talk) 18:34, 3 January 2013 (UTC)

historical quote
Miller (1987) cited in Hillinger (2004): "…Historically, economists have contributed at least as much as political scientists to the pure theory of voting. The theory of voting has its origins in the work of such enlightenment philosophers and mathematicians as Borda, Condorcet and Laplace. Little further progress was made until some forty years ago when the economist Duncan Black wrote a series of articles (most notably Black, 1948) on the logic of committees and elections, which were subsequently consolidated into a book (Black, 1958). Since Black revived the subject, a number of economists and political scientists have made important contributions. Indeed, the theory of voting has to some extent been subsumed by the more recent and abstract theory of social choice, which was virtually invented by the economist Kenneth Arrow (1951)."

Would make a great citation for the history section. Also, we should definitely be mentioning Black. Homunq (࿓) 13:33, 1 January 2013 (UTC)

Strategic Condorcet Criterion Citation Needed - Concern Over Hypothetical Conditions In Table
~Q: Is there a citation for "Strategic Condorcet"?

~Q: Are there citations for how to apply this criterion, which involves various hypothetical conditions and suppositions (i.e. predicting game theory outcomes), to each of the voting methods?

~Q: I wonder, if IRV voters have perfect information, how can we be sure that the "Push Over" effect would not be avoided, by voters compromise voting and hence correctly revealing the Condorcet winner, just as approval voters may compromise vote based on predicting how other might vote?

The problem I see raised by the last question, is that it places the determination into the realm of qualitative, if you will, rather than mathematical compliance terms.

I am sympathetic to those who want to show Approvals strong theoretical Condorcet performance under Nash Equilibrium etc., however I am very concerned that inventing a criterion is not the best way to express this strength.

Even so, I'm not sure that on the general voting page, where we have so many criteria that we should open up the possibility of criteria performance under various hypothetical conditions.

I strongly suggest that we include only mathematical compliance criteria, not hypothetical conditions.

At minimum, I would request that this criterion and Condorcet performance of Approval voting be moved to the Approval Voting page, a worthy topic, until it has been better vetted or given citations.

Filingpro (talk) 01:48, 20 May 2013 (UTC)

Citation For IIA and Rating Methods Compliance? + Question
Q1: Is there a citation for supposedly “passing” Independence of Irrelevant Alternatives by rating methods such as Approval in the compliance table?

Q2:

GIVEN: In plurality voting if A is elected over B in a two-candidate race, then adding spoiler C to the ballot can split votes for A, by some voters changing their ballots from A to C, causing B to win (hence violating IIA).

As with the example above, how is it not the case that in an approval voting where A is elected over B in a two-candidate race, then adding spoiler C to the ballot can split votes for A, by some voters changing their ballots from A to C ("bullet voting"), causing B to win (hence violating IIA)?

Filingpro (talk) 01:44, 23 May 2013 (UTC)

Does IRV Satisfy "Strategic Majority" Condorcet, Participation, Monotonicity, and LIIA? + PROPOSAL BEFORE EDIT
Q: In the compliance chart under "Strategic Majority Condorcet", where are the citations or argumentation for each of the method's stated compliance, except for Approval?

Q: If all voters have perfect information about other voters preferences and vote strategically, and it is assumed voter preference listings will not be truncated (i.e. guaranteed majority participation as stated in the "criterion"), then does IRV satisfy Condorcet, Participation, Monotonicity, and IIA?

EXAMPLE 1

Consider how the Condorcet winner is lost in IRV due to the "Push Over" candidate advancing to the final round, while the Condorcet winner suffers from early elimination:

Voter preferences (sincere):

28: R > C

5: R > L

30: L > C

5: L > R

16: C > L

16: C > R

In IRV, L wins but C is Condorcet winner. But if the voters have perfect information about preferences and vote strategically, then "R > C" will compromise vote, and choose "C > R", electing C the Condorcet winner. Simply put, those who are voting for the "Push Over" candidate ("R") and prefer the Condorcet winner over the other alternates, just vote for the Condorcet Winner first instead, an obvious strategy in their interest, because the "Push Over" candidate has no chance of winning no matter what rational strategies any coalition might attempt. This is akin to Approval voting always electing a Condorcet winner when voters compromise vote with perfect strategy.

Note importantly in the above example there is no game of "chicken" - i.e. a situation where one coalition can gain from playing a non-cooperative hand, taking advantage of another who plays a cooperative hand. There is no such unstable situation which deters the voting method from electing the best cooperative candidate C, the Condorcet winner.

For example, consider each of the other coalitions, none of which have a rational strategy leading to advantage:

- Voters "R > L" can do nothing further to change the winner from L to R, since R is defeated head-to-head by both L and C, while R is already maximized on the ballot.

- Voters "L > ?" have no interest in changing the winner from L. There is also no rational winnable counter-strategy against the other voter's who avoid voting for the push over candidate R. If "L > ?" voters assume "R > C" does not vote strategically, then there is no strategic response necessary. If "L > ?" voters assume "R > C" voters do vote strategically (i.e. switch "C > R") then they are powerless to defeat C with a whopping majority rule (In fact, Monotonicity violations are most likely impossible under "strategic" IRV voting, at minimum for this example).

- Voters "C > ?" are in a position of strength because they support the candidate who beats all other candidates according to voters' sincere preferences. Therefore holding their first choice "C" will necessarily force "R > C" voters to support C first because it is in their interest. Meanwhile "R > C" voters know that "C > R" voters wont change to R first, because there is no strategic advantage. (It is also noteworthy that if "R > C" voters logically vote "C > R" having perfect information, then IRV passes Participation criterion.)

EXAMPLE 2

Consider another IRV example (unrelated to Monotonicity and Participation issues), when IRV eliminates a Condorcet winner prematurely, one having weaker plurality strength:

49: L > C

48: R > C

2: C > L

1: C > R

NOTE: for this example it is important to note that the measure for passing the "Strategic Majority Condorcet" assumes there are no ties in the strategy equilibrium - we are allowed to consider only "For a large electorate, if there is an equilibrium with no tie" (see the citation for approval voting's determination http://halshs.archives-ouvertes.fr/docs/00/12/17/51/PDF/stratapproval4.pdf). In other words, R's enthusiastic supporters have perfect information that L will defeat them in a head-to-head race, despite how close the race might be.

L is the IRV winner but C is the Condorcet winner. But if the voters have perfect information about preferences and vote strategically, then "R > C" will compromise vote, and choose "C > R", electing C the Condorcet winner. Simply put, those who are voting for the "Push Over" candidate ("R") and prefer the Condorcet winner over the other alternates, just vote for the Condorcet Winner first instead, an obvious strategy in their interest, because the "Push Over" candidate has no chance of winning no matter what rational strategies any coalition might attempt. Once again, this is akin to Approval voting always electing a Condorcet winner when voters compromise vote with perfect strategy.

These examples raise several important questions:


 * 1) Where are the citations and determinations for "Strategic Majority Condorcet" coming from? Are they correct? IRV appears to pass but "No" is put in the table.
 * 2) There is no citation in the literature for proposing this as a mathematical criterion, only a theorem and proof of how approval elects Condorcet under perfect strategy conditions.
 * 3) If we are allowed, as editor's to invent criterion to show how certain systems perform better, then should we not also add "Strategic Majority Participation" for which IRV passes, and also "Strategic Majority Monotonicity" for which IRV passes?
 * 4) The idea of how voting systems perform under strategy is a very worthy topic, but shouldn't there be an entire section on voting under strategy? Given we don't have citations in the literature for the various systems under "strategic" criteria, nor citations for the criteria, it seems highly problematic to list this as a criterion in the "Comparison" table, rather than discuss them in essay form.

PROPOSAL:

1) we eliminate "Strategic Majority" criteria from the table because the criterion is WP:OR and the determinations for important voting systems are WP:OR

2) we put the comments either under footnotes for "Condorcet", or some other section about how voting systems respond to strategy (i.e. how compliance might change), or probably best if it goes into the approval voting page since this is a particular theorem about approval, not a criterion

Otherwise, we would have to put blank white in the determinations for all the voting systems that don't have a citation. Likewise, any editors could invent other criteria like "Strategic Participation", "Strategic Monotonicity" for which IRV passes.

And finally, once again, I am EXTREMELY concerned that on the one hand we have added (or allowed) an entire compliance criterion for which the application to approval is mathematically derived from a model of voter's strict ordinal preferences, in order to present approval favorably. On the other hand, same editors claim that we can not apply Later-No-Harm to approval using voter's strict ordinal preferences, which portrays approval unfavorably.

To see how approval's mathematical determination for "Strategic Condorcet" is derived from voters strict ordinal preferences, please read the source provided http://halshs.archives-ouvertes.fr/docs/00/12/17/51/PDF/stratapproval4.pdf, see the section under "Model", which reads "...it turns out that the obtained results can all be phrased in terms of preferences. That is: It will be proven that rational behavior in the considered situation only depends on preferences." It is not surprising that the correct mathematical model of the subject matter at hand must include the voters preferences as part of the model, otherwise the results would be useless.

Filingpro (talk) 01:16, 27 May 2013 (UTC)
 * Thanks for a thoughtful discussion, Filingpro. While I think that the "strategic majority" criterion is significant, and as you note we do have a valid citation for Approval in this regard, I am forced to acknowledge that you're right; we don't have citations for other systems, so the majority of this column is OR. The column should be removed, and discussion of this issue should be moved to approval voting.
 * I apologize that this is the first time someone has noticed your valid complaints on this issue.
 * As for IIA (above talk section): I'm pretty sure I can find a citation for at least Range voting, and one which uses the "no changing ballots allowed" definition of the criterion which MJ trivially passes. I'll look around a bit. Please do not remove the column in the meantime. Homunq (࿓) 15:12, 27 May 2013 (UTC)
 * Obviously, the example involving IRV is OR and not usable on the page. However, I think you're right: a majority condorcet winner is always the unique strong Nash equilibrium for IRV. Still, I believe you could construct cases where there is a non-majority Condorcet winner (ie, some voters are indifferent between that winner and some other candidate) and yet that candidate is not a strong Nash equilibrium winner for IRV. For Range and MJ, on the other hand, that non-majority CW candidate will still always be a strong Nash equilibrium winner, just not a unique one. This is a giant pile of OR and so if you want to discuss it further, we should probably take it to my talk page (or even email; you can use "email this user" for me and I'll respond.)
 * ps. the criteria as currently on the page reads (strategic (majority Condorcet)), not ((strategic majority) Condorcet). As you correctly point out, IRV passes that; but my claim here is that you could define a "((non-unique strategic) Condorcet)" criterion which would be passed by Range and MJ but not by IRV. I repeat though, it's a buncha WP:OR, so I don't plan to put anything like that on the page. Homunq (࿓) 15:31, 27 May 2013 (UTC)


 * Thank you, Homunq. Yes I do think the performance of these voting systems under strategy is very important to understand and yes I am very interested in the distinction you make regarding how they perform differently considering when voters are indifferent, and whether equilibrium winner must be unique. On a cursory level, when I poked around with approval and the Nash equilibrium concept, I found the most likely scenarios for failure involved some voter indifference, close ties, or incomplete voter preference rankings prior to translation to the approval ballot, although I cannot be sure a failure was found. So yes, I am quite interested in exploring this topic further.


 * I would like to defer to you first for the edits, because I think you have more familiarity with regard to the evolution of this column and its various footnotes, so that you are probably more qualified to surgically remove it while still respecting contributions from past edits etc. However, if you would rather defer this to me I will rise to the occasion.


 * One possible suggestion is I notice there is a duplicate, general footnote that now appears in many of the compliance for the systems under Condorcet criterion, something like “Condorcet…is incompatible with…etc. etc.”. Since this comment has no specific wording for a particular method – but to all those that fail, I wonder if we could put it in the rollover for the criterion itself. For example, it might be useful for the reader to be able to scan the top of the table and see how the different criteria are incompatible with each other. This frees up the footnote for adding the specific strategy performance citations for specific voting systems such as approval in approval’s box.
 * Filingpro (talk) 19:15, 27 May 2013 (UTC)

Cumulative Voting
The summary table does not contain a line about Cumulative Voting. Can someone please add it? Thanks for all the hard work! --Erel Segal (talk) 14:42, 13 September 2013 (UTC)

Additionally, the text here says that "It is considered a proportional system in allowing a united coalition representing a m/(n+1) fraction of the voters to be guaranteed to elect m seats of an n-seat election", but, this is not compatible to the text in Cumulative Voting. Can someone please explain, how Cumulative Voting allows a coalition representing a m/(n+1) fraction of the voters to guarantee m seats of an n-seat election? --Erel Segal (talk) 18:58, 14 September 2013 (UTC)

LNH & Approval – No Citation Needed
In Woodall's Later-No-Harm, if b is a “later preference” than a, then a>b, given.

The objection to applying Later-No-Harm to Approval is really an objection to applying ordinals to Approval in general.

But ordinals do apply to Approval because a>b can be approvals ab, a, or none, axiomatically. So what’s the objection? I think objectors, if there are any, have an obligation to explain.

Note1: Saying LNH requires a “later” preference on an Approval ballot is a false premise, because Woodall doesn’t use an Approval ballot, he uses ordinals. We have to explain why ordinals can’t be applied which violates what is axiomatic.

Note2: Saying we need a citation would be disingenuous because taking what is given and applying an axiom requires no citation. For example, given the numbers 1 and 2, we don’t need a wiki citation to state 1 + 2 = 3 by axioms of addition. This is not WIKI SYNTH.

I realize this has already been discussed and a compromise agreed to, but I believe we overlooked the definitions in Woodall's document which explicitly refer to the voter's preferences, not the literal markings on an Approval ballot. Thanks for everyone’s consideration.

Filingpro (talk) 21:28, 11 October 2013 (UTC)


 * Yes, it has been discussed, and a compromise was reached. No, I don't think your argument here is anything new relative to that discussion. Please, let it lie, or at least find a sock puppet or something. (joke) Homunq (࿓) 00:14, 27 October 2013 (UTC)

IIA Imprecise Stipulation: "Votes for other candidates can not change" - Depends On Voter Cutoff Model Assumed
The article currently reads:

"Independence of Irrelevant Alternatives (IIA)— Does the outcome never change if a non-winning candidate is added or removed (assuming votes regarding the other candidates are unchanged)?[4] For instance, plurality rule fails IIA; adding a candidate X can cause the winner to change from W to Y even though Y receives no more votes than before."

The problem I see: We changed some votes from W to X

In IIA, if we stipulate that the markings on the ballots for the other candidates cannot change when we add or remove an irrelevant candidate, then Plurality would pass IIA but it does not, which leads to a contradiction.

CASE 1:

A 51

B 49

A Plurality winner

If we introduce candidate X preferred by A voters which is irrelevant to the race between A and B, then no A voters can change their vote to X without withdrawing their votes for A. If we can not change votes for A we can not change the winner from A to B.

CASE 2:

X 3 (Voters prefer X > A > B)

A 48

B 49

B Plurality winner

If we remove candidate X from the election which is irrelevant to the race between A and B, then no former X voters can change their votes from X to A without changing their votes for A. If we can not add votes for A we can not change the winner from B to A.

This is incidentally why all non-dictatorial and Pareto efficient voting systems including Plurality, Approval, and Range fail IIA, because ballot markings may change when a candidate is added or removed based on the voter cutoff model assumed, while voters' ordinal and relative weighted preferences stay the same.

For example, in Approval voting, if all voters prefer X > A > B and approve only X, but then X is removed, then all voters prefer A > B so A must win to satisfy Pareto Efficiency, but the criterion is failed unless at least one voter can approve A.

Approval dictatorship:

999999 X (Voters prefer X > A > B)

1 B

Remove X from the election. If we adopt a mathematical model such that the approval cutoff for X voters can not change, then the B voter is the dictator.

Filingpro (talk) 08:24, 27 October 2013 (UTC)

Removal of Condorcet-IRV method
Why was the Condorcet-IRV method removed from the article? Wat 20 16:03, 11 November 2013 (UTC)

Add Symmetrical ICT?
I found another voting system which I think complies with the Condorcet Criterion and is designed to also comply with the two criteria mentioned above, Favorite Betrayal and Chicken Dilemma. I don't see it mentioned on Wikipedia, so maybe somebody should add a page for it and analyze whether or not it meets the criteria in the table on this page.

http://wiki.electorama.com/wiki/Symmetrical_ICT

Myrkron (talk) 21:03, 5 November 2013 (UTC)


 * Again, those are not widely accepted criteria. Markus Schulze 08:30, 6 November 2013 (UTC)


 * Qaanol (talk) 02:47, 2 February 2014 (UTC)

Additional Criteria?
When researching voting systems on the web, I've come across two more interesting criteria. Are they excluded on purpose?

Favorite Betrayal: If the winner is a candidate who is top-voted by you, then moving an additional candidate to top on your ballot shouldn't change the winner to a candidate who is not then top-voted by you.

Chicken Dilemma: http://wiki.electorama.com/wiki/Symmetrical_ICT Basically, you shouldn't be able to make your first choice win by rearranging your lower choices.

Myrkron (talk) 21:04, 5 November 2013 (UTC)


 * Those are not widely accepted criteria. Markus Schulze 08:30, 6 November 2013 (UTC)


 * Not excluded on purpose, but they appeared only recently, and long after the voting system article was created. But I would point out that these criteria are very similar to IIA and are all incompatible with Condorcet criterion. Wat 20 02:34 12 November 2013 (UTC)


 * The favorite-betrayal criterion has certainly been known for many years. It had a Wikipedia article which seems to have been deleted just last summer. Here is the most recent Wayback Machine snapshot of it, from June, 2013.
 * Moreover, this very Voting Systems page had the favorite-betrayal criterion included in its table of compliance for nearly all of the last two years, from January, 2012, through October, 2013.
 * Additionally, and I know this is not at all scientific, a Google search for ("favorite betrayal criterion" -site:wikipedia.org) yields 45,300 results, while ("reversal symmetry criterion" -site:wikipedia.org) yields only 2,360. That is over 18 times more non-Wikipedia hits for favorite betrayal criterion than reversal symmetry criterion.
 * So I propose we reintroduce FBC into the discussion and table of voting system compliance. Qaanol (talk) 01:27, 2 February 2014 (UTC)


 * For "favorite betrayal criterion", there are no hits in Google Books. And there are only 4 hits from 3 different authors in Google Scholar. Markus Schulze 15:08, 2 February 2014 (UTC)


 * Just as a note on books: "Gaming the Vote" by William Poundstone does mention the FBC, but it calls it a "property", not a "criterion". Homunq (࿓) 20:38, 20 March 2014 (UTC)

Changes made to citations
I changed a few citations formats, and replaced the shortened citations with the full citations. I also used named references to make those changes. This is the link to my sandbox. https://en.wikipedia.org/wiki/User:Lisax31/sandbox1 I am looking forward to your suggestions and feedbacks. Thanks! Lisax31 (talk) 04:13, 7 March 2014 (UTC)
 * Unrelated topic: I deleted the section that I myself created. I did this after I realized that what I said did not make sense. Sorry - oops - Boyd Reimer (talk) 16:50, 22 March 2014 (UTC)

Blacklisted Links Found on the Main Page
Cyberbot II has detected that page contains external links that have either been globally or locally blacklisted. Links tend to be blacklisted because they have a history of being spammed, or are highly innappropriate for Wikipedia. This, however, doesn't necessarily mean it's spam, or not a good link. If the link is a good link, you may wish to request whitelisting by going to the request page for whitelisting. If you feel the link being caught by the blacklist is a false positive, or no longer needed on the blacklist, you may request the regex be removed or altered at the blacklist request page. If the link is blacklisted globally and you feel the above applies you may request to whitelist it using the before mentioned request page, or request its removal, or alteration, at the request page on meta. When requesting whitelisting, be sure to supply the link to be whitelisted and wrap the link in nowiki tags. The whitelisting process can take its time so once a request has been filled out, you may set the invisible parameter on the tag to true. Please be aware that the bot will replace removed tags, and will remove misplaced tags regularly.

Below is a list of links that were found on the main page:


 * http://halshs.archives-ouvertes.fr/docs/00/12/17/51/PDF/stratapproval4.pdf
 * Triggered by  on the global blacklist

If you would like me to provide more information on the talk page, contact User:Cyberpower678 and ask him to program me with more info.

From your friendly hard working bot.— cyberbot II NotifyOnline 15:28, 27 March 2014 (UTC)

✅ This issue has been resolved, and I have therefore removed the tag, if not already done. No further action is necessary.— cyberbot II NotifyOnline 07:28, 29 March 2014 (UTC)

Deletion review on Favorite betrayal criterion
The favorite betrayal criterion page has been deleted in accordance with Wikipedia policy; that deletion is now undergoing a review. Your comments are welcome. Homunq (࿓) 14:33, 14 April 2014 (UTC)

RFC - favorite betrayal?
Should the favorite betrayal criterion (FBC) be included in the criteria table on this article?

Survey

 * Support as RFC originator. This criterion has adequate sourcing and is sufficiently distinct from the other criteria in the table to merit inclusion. In general, this article is improved by including the various criteria that are discussed in regards to voting systems, and one more criterion in the context of many others is not WP:UNDUE. Homunq (࿓) 13:23, 19 April 2014 (UTC) (UNDUE is the issue here, not notability. As WP:GNG clearly states, "If a topic does not meet these criteria but still has some verifiable facts, it might be useful to discuss it within another article.") Homunq (࿓) 15:18, 19 April 2014 (UTC)
 * I have not reviewed the content and sources in question so as to voice an opinion on this particular case just yet, but based on the above comment, it bears noting here that notability is not a factor relevant to GNG alone. Snow (talk) 00:51, 21 April 2014 (UTC)
 * Fair enough. Of course, I also claim it's notable; YMMV. Homunq (࿓) 01:03, 21 April 2014 (UTC)
 * Yes, I was just looking at that discussion. And a number of the other numerous discussions on the topic.  Seems to have been going on for some time now, probably longer than necessary, but marks to everyone for keeping their cool and staying courteous and on the point of policy (that I've seen so far, anyway).  Still, it's a good thing S Marshall decided to call for a larger (and hopefully more stabilizing) consensus.  Getting back to the issues though, I wonder if I could impose upon you to list, somewhere above or bellow in this discussion, exactly the sources you would like to include in the present version of this article?  I'm sure we can all find the version with the content/claims under discussion, but just to make sure I'm not missing something obvious, wouldn't mind having the sources listed briefly here for reference. Snow (talk) 02:55, 21 April 2014 (UTC)
 * Response below. Homunq (࿓) 14:55, 21 April 2014 (UTC)


 * Oppose The "favorite betrayal criterion" doesn't meet Wikipedia's notability criteria. See this discussion and this discussion. Markus Schulze 14:56, 19 April 2014 (UTC)
 * notability rules do not apply in this case. What matters here is the quality of the references.  If there is a controversy within the references that should be spelled out.  The debate over what is correct has no business in this talk page.  What we should be discussing is what the most respected literature says.  Homunq seems to have supplied references, and I think your position should either be that the references are inadequate, or there is a controversy.  I see no place for the OR that seems to be playing out on the talk page.Bob the goodwin (talk) 11:10, 24 April 2014 (UTC)


 * Oppose The FBC refers to how a particular "voting strategy" is handled, rather than a voting criterion per se for this article - and is like all "voting strategies" not directly related to differences in "voting systems." Other strategies include "bullet voting" and "wrong party voting in primaries" etc. None of which are in this article.  See Tactical_voting Collect (talk) 22:09, 20 April 2014 (UTC)
 * Huh? This RfC is about the favorite betrayal criterion, not the phenomenon/strategy of favorite betrayal in general. I think that even Schulze wouldn't deny that the FBC is a well-defined criterion which is passed by some voting systems and failed by others — exactly like all the other criteria in this article. What we have to decide here is whether it merits mention here. Perhaps this alternate version of the table would help clarify this point. Homunq (࿓) 22:50, 20 April 2014 (UTC)
 * I read the article. I looked up papers.  The "criterion" is with reference to a particular voting strategy which has been postulated (but not shown to have had any wide usage as it requires fairly precise knowledge of where the vote percentages will end up).   It is not, however, directly relevant to this article, but I suggest it might be relevant to the "voting strategy" article.  RfCs are "Requests for Comment" and not "Requests for Agreement" by the way. Collect (talk) 23:47, 20 April 2014 (UTC)
 * Thanks for the clarification. I understand your point now and it's certainly appropriate to the question for you to talk about relevancy. I misunderstood you above and thought that you were denying that FBC was even a well-formed criterion. Sorry. Homunq (࿓) 00:15, 21 April 2014 (UTC)


 * Support because favorite betrayal is the strongest strategy against Condorcet and approval methods. EllenCT (talk) 23:56, 20 April 2014 (UTC)
 * I would rather say that burying and compromising are the strongest strategies under Condorcet and approval methods. Markus Schulze 07:07, 21 April 2014 (UTC)
 * If we don't have a reliable source which says that, then it could be considered original research (which would be a problem). Balaenoptera musculus (talk) 14:13, 25 April 2014 (UTC)


 * Support as per Homunq - article has due weight and may be useful for readers Drowninginlimbo (talk) 23:59, 20 April 2014 (UTC)
 * Comment Homunq writes: "I think that even Schulze wouldn't deny that the FBC is a well-defined criterion." Actually, I do deny that the favorite betrayal criterion is well defined.
 * I have made the observation that, whenever I ask one of the few supporters of this criterion a non-trivial question about this criterion, he modifies the definition of this criterion a little bit so that this definition answers this question, but simultaneously opens many new questions.
 * The current version of the "favorite betrayal criterion" article says: "A voting system satisfies the favorite betrayal criterion if there do not exist situations where a voter is only able to obtain a more preferred outcome (i.e. the election of a candidate that he or she prefers to the current winner) by insincerely listing another candidate ahead of his or her sincere favorite."
 * Example: Suppose in a concrete situation a concrete voter can, when voting system X is used, obtain a more preferred outcome only by insincerely listing another candidate ahead of his sincere favorite. Suppose in the same situation the same voter can, when some other voting system Y is used, obtain the same more preferred outcome by insincerely listing another candidate ahead of his sincere favorite and also by some other strategy that doesn't require that he lists another candidate ahead of his sincere favorite. Then, voting system X violates the favorite betrayal criterion and voting system Y satisfies the favorite betrayal criterion. So, according to the favorite betrayal criterion, voting system Y is better than voting system X. But this is not only in drastic contrast to the consensus in social sciences, it is the exact opposite of the consensus in social sciences, which says that a voting system is the better the less manipulable it is. The consensus in social sciences would be that voting system X is better than voting system Y because in voting system X the outcome can be manipulated only by insincerely listing another candidate ahead of one's sincere favorite.
 * When you promote an idea that is in drastic contrast to the consensus in your academic field, you need good sources that support your idea. But there is not a single reliable source that focuses on the favorite betrayal criterion. Markus Schulze 06:04, 21 April 2014 (UTC)
 * You're claiming 3 things here: ① the FBC is not well-defined; ② whenever you bring up an (unspecified) objection, (unspecified) FBC supporters make an (unspecified) alteration to their definition; ③ the FBC contradicts a "consensus in social sciences" due to the example you give.
 * Your argument is structured as if ① were the main claim, and ② and ③ were evidence. But ② is too vague to constitute evidence for anything; and ③ actually contradicts ①, because in order to construct the example in ③, you use a clear definition to decide whether your hypothetical systems X and Y pass the FBC. In fact, it's because I've seen you make arguments like ③ before that I didn't expect you to believe ①.
 * Furthermore, I'd say that your argument in ③ is wrong for three different reasons:
 * Ⓐ: I do agree that if, in an example like yours, X were not manipulable in a certain scenario, while Y were manipulable, the "social science consensus" would be that X was better. But in your actual example, X is already manipulable by dishonest strategy. I think many social scientists would say it's already maximally broken in that case, and Y is not "more broken" by also being manipulable through semi-honest strategy.
 * Ⓑ: Since Condorcet and Arrow, one of the main themes of voting theory is that desirable criteria often contradict each other. So when you say that the FBC is an undesirable criterion because it contradicts your claimed consensus preference for "less manipulable" systems as you've defined them, someone else might just as well claim that non-dictatorship is undesirable because it contradicts the consensus preference for independence of irrelevant alternatives. I'd call that other claim silly, and I think yours is no better.
 * Ⓒ: Insofar as the FBC does did run contrary (as you claim; or orthogonal, as I claim) to other commonly-held desiderata, that makes would make it more, not less, notable. The point of the criteria table in this article should be to give a reasonably balanced view of the different reasons one might prefer different systems. This is an encyclopedia, not a persuasive essay marshalling arguments to support a single point of view or one person's preferred voting system.
 * Homunq (࿓) 13:04, 21 April 2014 (UTC)
 * The favorite betrayal criterion is not a criterion that says that a voting system should not be vulnerable to a certain kind of strategies. It is a criterion that says that a certain kind of strategies should be available and working. This is in drastic contrast to the consensus in social sciences, because this consensus says that it is desirable when strategies don't work.
 * Homunq wrote: "Insofar as the FBC does run contrary to other commonly-held desiderata, that makes it more, not less, notable." No, the more a claim is in contradiction to the academic consensus the more it needs reliable sources to support its notability. Markus Schulze 14:56, 21 April 2014 (UTC)
 * Schulze's allegation that there's a consensus is clearly WP:SYNTH. I've already explained why I disagree with it. Homunq (࿓) 15:13, 21 April 2014 (UTC)


 * My view is that FBC should be included in the table here but should not have a separate article. I'm agnostic about whether FBC is notable, but I'm of the view that it's probably not notable enough to warrant a whole article all of its own.  I think the FBC page should redirect to this page.  I think there should be a one- or two-sentence summary of FBC on this page and it should appear in the table.  NB: Even if it isn't notable, FBC could still be included here.  Notability is about whether a topic should have its own article.  Notability is not normally relevant to the content of any specific page.— S Marshall  T/C 00:19, 22 April 2014 (UTC)


 * Oppose The sources listed below in 'response to snow' don't seem strong enough to me. WP:SCHOLARSHIP prescribes secondary sources or (failing that) peer-reviewed papers. I'm not seeing any of those in the list below. Feel free to correct me if I've missed them. As it doesn't seem to have strong enough reliable sources, including it seems like undue weight to me. Of course, I'd reconsider this view if stronger sources emerge. Balaenoptera musculus (talk) 14:11, 25 April 2014 (UTC)
 * Both of the book references (Gaming the Vote: Why Elections Aren't Fair (and What We Can Do About It) by William Poundstone ISBN-10: 0809048922, p229; and ), and the peer-reviewed paper talking about the more general phenomenon of favorite betrayal , attest to the existence of this criterion. These meet your threshold. The sources that I listed below are intended to show the more specific properties. The difficulty is that the most reliable sources are not the same as the most detailed ones. Note that this is true of several of the other criteria (though I won't say which, because I don't want to make this page worse just to make a WP:POINT.) Homunq (࿓) 17:48, 25 April 2014 (UTC)


 * Support Thanks for the info, Homunq. In my view the following is a WP:RS which mentions the existence of the FBC, so its inclusion in the article is justified. "Using and Understanding Mathematics: A Quantitative Reasoning Approach, 5th Edition / By Jeffrey O. Bennett, William L. Briggs / Publisher: Pearson / ISBN-13: 978-0-321-65279-9" p.650 quote "53. Other Fairness Criteria / The fairness criteria discussed in this unit are not the only ones. Investigate other criteria (such as the favorite-betrayal criterion, the strong adverse results criticism and the weak defensive strategy criterion). Discuss the merits of these criteria and their shortcomings." I haven't checked out the other references in any detail yet. Balaenoptera musculus (talk) 14:44, 28 April 2014 (UTC)


 * Support, as per S Marshall, and because Wikipedia should err on the side of inclusiveness of information included. I note a strong feeling that Markus Shulze is motivated by a COI. Homunq's COI is unclear. I disagree that "peer reviewed publication" is a minimum standard for inclusion. More important is whether independent others have discussed it. I have limited sympathy for Markus' view that FBC is unreliably defined, however that needs to be taken with a grain of salt given that the Schulze Method apparently fails FBC. --SmokeyJoe (talk) 05:23, 27 April 2014 (UTC)
 * Homunq is permanently claiming that I had a COI and that he hadn't a COI. However, it is a matter of fact that, outside Wikipedia, Homunq is an ardent supporter of majority judgment which, according to Homunq, satisfies FBC. The difference between Homunq and me is that I am honest about my true identity (in so far as I am not denying that I am promoting the Schulze single-winner election method) while Homunq is making false claims about his true identity (in so far as he is denying that the fact, that he is promoting a concrete single-winner election method, could be a reason for a COI). Markus Schulze 06:43, 27 April 2014 (UTC)
 * Majority judgment looks to sit ill with WP:NOR, and appears to be written from the one primary source.
 * I'd like to see FBC mentioned, but maybe the big table of compliances needs to be divided into well recognized criteria, and other criteria. --SmokeyJoe (talk) 07:25, 27 April 2014 (UTC)
 * Schulze makes several errors here. (I'd also like to remind him of WP:OUTING. The above is not a violation, but for instance getting specific about my leadership positions in voting advocacy organizations would be. However, if any editors here are curious about my real identity, just email me using the "email this user" link to the left of my talk page, and I'd be happy to reveal/discuss it in private. Or, to be honest, I'm sure you could find me by googling a few terms from below.)
 * First, I have been scrupulous about WP:AGF, and have not accused Schulze of violating the WP:COI policy. I realize that this is somewhat hairsplitting, as I have spoken of the possible other interests he may be pursuing here (though, if I'm not mistaken, I'm actually not the first editor to have brought this issue up). But just as Schulze insists just below, precision is important here.
 * Second: while Schulze has correctly guessed my off-wikipedia identity, he is wrong to call me an "ardent supporter of majority judgment". I consider MJ to be an excellent single-winner voting system, but as a matter of practical advocacy I prefer approval voting, while as a matter of theoretical ideals I prefer Simple Optionally-Delegated Approval or Majority Approval Voting. (Both of the latter are my own inventions. SODA shares almost nothing with MJ, but MAV is a minor variation on the MJ idea).
 * Third: Schulze accuses me of making "false claims about my own identity". In his rush to quoque, he seems to have momentarily forgotten about WP:CIVIL, but that's OK. The fact is that I have openly admitted that I have a motive here; and, as I say above, I'd be happy to go into further detail in private with anyone who's interested. But this is not a symmetrical situation; there is a difference between wanting something to be discussed in Wikipedia because I think it is important, interesting, and verifiable, and not wanting it discussed because it makes his system look bad.
 * Finally: this is not really the place to discuss the majority judgment article, but I dispute the characterization of the book by that name as a "primary source" in the Wikipedia-accepted sense. Homunq (࿓) 14:16, 27 April 2014 (UTC)
 * In the thesis by Hughart, the definition of FBC is simply false; he mixes up FBC and "compromising". The same can be said about the paper by Stensholt. Also, in several papers by Ossipoff and Smith, FBC is defined in such a vague manner that the difference between FBC and "compromising" doesn't become clear; usually, they use phrases like: "A voter should never be forced to insincerely list another candidate ahead of his sincere favorite." However, this could also be a definition for "compromising". Even that paper that Homunq considers to be the best, doesn't contain a clear definition for FBC. Even the definition proposed by Homunq ("Can voters be sure that they do not need to rank any other candidate above their favorite in order to obtain a result they prefer?"), doesn't make it clear that FBC and "compromising" are different things. Markus Schulze 07:55, 27 April 2014 (UTC)
 * FBC and compromising do sound the same. A mention and citation of FBC at Tactical voting might be enough?  --SmokeyJoe (talk) 08:17, 27 April 2014 (UTC)
 * FBC and "compromising" do sound the same, but they are not the same. FBC supporters frequently try to mislead the readers into believing that FBC and "compromising" were the same. They try to portray FBC as if it was "compromising". Markus Schulze 08:51, 27 April 2014 (UTC)
 * "Invulnerability to compromising" says that there should never be a situation where it is advantageous to rank another candidate ahead of one's sincere favorite. FBC says that, whenever it is advantageous to rank another candidate ahead of one's sincere favorite, there should also be an available strategy that doesn't require that this voter ranks another candidate ahead of his sincere favorite. Markus Schulze 11:28, 27 April 2014 (UTC)
 * These things are clearly the same subtopic. FBC can get a mention at Tactical voting.  This mention can mention that FBC is imprecisely defined and not the same as whatever.  If FBC is mentioned there, then there is less need for it to be mentioned here.  --SmokeyJoe (talk) 11:51, 27 April 2014 (UTC)
 * It's interesting that Schulze says that Hughart's definition is "simply false"; that implies that he accepts that there is enough logical clarity in the concept to decide whether a definition is or isn't false. I agree that Hughart does confuse the issue a bit, but in his discussion, not in his definition. But again, the fact that we can largely agree here shows that there is a clear, agreed-upon concept here, and that's exactly what I'm arguing deserves coverage on this page. Schulze has a sophisticated understanding of many voting system criteria including the FBC; why would he want to deny others the opportunity to develop the same?
 * (As for the Stensholt paper: there's no clear definition there, and it's very much in a context of strictly ordinal systems, which cannot pass FBC, so sure, I guess I'll agree on that point too.)
 * I'd be entirely amenable to a sentence clarifying the distinction between FBC and compromising, and think Schulze has a good point on that score.
 * I do want the FBC to be mentioned in this article and not just at tactical voting. The table here is the best source on Wikipedia, and one of the best sources on the internet, for a simple introduction to voting criteria.Homunq (࿓) 14:16, 27 April 2014 (UTC)
 * Seperately: Herr Doktor Schulze, you've now responded to this RfC many times, without making any commitment to abide by its resolution. Could you please comment on your position on this matter on the relevant DRN? Homunq (࿓) 14:22, 27 April 2014 (UTC)


 * Comment Dear Homunq, when we subtract all those papers that don't contain a definition for FBC and when we subtract all those papers where the definition for FBC is so vague that the difference between FBC and "compromising" doesn't become clear, which papers remain? Markus Schulze 13:01, 28 April 2014 (UTC)
 * I'd be happy to answer that question, if you'll answer mine: will you abide by the outcome of this RfC, whatever it is? If you want to include caveats, that's fine; I just want some mutual commitment, so we can avoid a (polite, slow-mo) edit war.
 * I'll even offer something further: I'd ask whoever is closing this RfC to ignore my own !vote if necessary to find consensus, if you do the same. I think it's important that this RfC be resolved as decisively as possible, and I am willing to explore such mutual stand-downs if you are. Homunq (࿓) 14:58, 28 April 2014 (UTC)
 * By the way: nice edit on the main article. I entirely endorse your reversion of the "criteria counts". Including such a count would be an open inducement to edit war about criteria inclusions; exactly the kind of thing I'm trying to avoid with this RfC. This underscores the point that you (Schulze) are absolutely a healthy influence for the page overall, as much as I disagree with you on the one issue of FBC. Homunq (࿓) 15:10, 28 April 2014 (UTC)

Discussion
To see references on the FBC, see Deletion_review/Log/2014_April_14 (as the older deletion processes linked from there). These references show that the FBC easily clears the hurdle of WP:V so this debate is about WP:UNDUE. Or, in plain English: nobody disputes the fact that the term "favorite betrayal criterion" refers to something with certain undisputed mathematical properties and a bearing on voting theory; the only question is whether it merits mention in this article. I believe that in cases like this (as long as, as in the current case, there are no other concerns such as WP:BLP) the presumption should be on the side of inclusion.

The reason this merits an RFC, rather than the regular editing process, is that it's reasonable to guess that this would lead to an edit war without an RfC. The Favorite betrayal criterion article has been the subject of an extraordinary 12 processes related to deletion and undeletion, one of which is ongoing at the moment. In recent years, these processes have been spearheaded by two users — User:MarkusSchulze and myself. I am convinced that both of us have acted in good faith in these processes, and that part of the reason that issue has been hard to resolve is that it is a genuine borderline case. However, I think it also bears mention that Schulze is the originator of the Schulze method, and that excluding the FBC from this page here makes his method look better. That's not a clear-cut WP:COI, but it does make me view his arguments with a certain extra skepticism. Homunq (࿓) 13:23, 19 April 2014 (UTC)
 * (Note: I believe that, however this RfC is resolved, it will take the pressure off of the deletion/undeletion debate, and, like a beautiful layer of nacre, prevent that debate from being a perpetual grain of sand in the Wikipedian oyster.) Homunq (࿓) 14:17, 19 April 2014 (UTC)

Questions Balaenoptera musculus (talk) 17:10, 19 April 2014 (UTC)
 * Given the history, won't this RfC just get appealed, through one process or another, by whichever of the involved editors considers themselves to have 'lost'? (and then that appeal get appealed, and so on and so on for round after round).
 * If this is only going to be the first round of an everlasting Kafkaesque bureaucratic rivalry then is it worth the time of independent editors getting fully acquainted with the issues, arguments and history in this case?
 * I for one will abide by this RfC in letter and spirit even if it goes "against" me, and would trust Schulze if he made a similar commitment.
 * Caveats that should go without saying:
 * If the decision is "no", then a new source appears that is unequivocally better than all current sources, such as a peer-reviewed paper with a significant focus on this criterion, I would consider the matter ripe for re-evaluation.
 * Obviously, I hope that my reasonableness won't be used against me (ie, I don't want people to argue "let's resolve against homunq simply because he's promised to shut up and Schulze hasn't." But if they don't specifically say that, I won't worry.)
 * Homunq (࿓) 17:44, 19 April 2014 (UTC)
 * Also, of course, the more people participate in this RfC now, the less likely it is that any future wikilawyering from either side will be able to overturn the result. Homunq (࿓) 12:01, 20 April 2014 (UTC)

More comments
 * I am a bit confused as to *exactly* what is being proposed here, but I did read up on the issue of "Favorite Betrayal" and I think it definitely deserves to be discussed somewhere on Wikipedia. 173.160.49.206 (talk) 15:51, 20 April 2014 (UTC)
 * I believe my suggestion that it fits best under Tactical voting as representing material directly relevant thereto is germane here. Collect (talk) 12:11, 21 April 2014 (UTC)

Response to Snow Rise
 * Snow rise asks above: "exactly [which] sources [would you] like to include in the present version of this article?"
 * This question is not the same as listing all the sources which help make the case for notability. For instance, this homework problem which asks the student to "investigate other criteria (such as Favorite Betrayal Criterion...)" is a strong argument that this is notable; if Wikipedia isn't for investigation homework, what is it good for? But a citation which says nothing but "please investigate" would of course be useless in the article.
 * So, to answer:
 * For the definition of FBC, I'd cite this thesis.
 * For the compliances of various systems with FBC, it's a little trickier. I'd like to note that there are several cases in the table where compliance citations are a bit weak; since we're dealing with objective mathematical facts, most editors here (including Schulze) have been willing to live with that, and I absolutely DO NOT want this RfC to result in somebody making a WP:POINT which upsets that understanding.
 * There are two kinds of citations we could use.
 * First is this work by Ossipoff and Smith. Obviously, that kind of material would not usually count as WP:V; it's self-published, and obviously so. However, I would note that that very page has been cited twice (and, somewhat generously, characterized as a "preprint") in peer-reviewed work by Steven Brams; so, bloggish and sketchy as it is, it arguably could be considered part of the academic literature. Certainly, its relevance to the questions of compliance is clearest of all the possible citations.
 * Second are other, more apparently "legitimate" preprints of work by Smith:    (this last one has been "cited by 35")  ("cited by 5"); or by others . These have not been published in peer-reviewed journals, but in my view they represent solid academic work, and since several have gotten citations, it seems my opinion is shared by some. (Having interacted with Smith, I'd guess that it's his often-abrasive personality that has kept his work from publication, not any defects in his logic. But of course, that's my own OR.)
 * Of these, I'd guess that the "cited by 5" [paper is probably the best citation for the article.
 * Cheers, Homunq (࿓) 14:52, 21 April 2014 (UTC)
 * Another (probably inferior) possibility would be Endriss, U. “Vote Manipulation in the Presence of Multiple Sincere Ballots.” In Proceedings of the 11th Conference on Theoretical Aspects of Rationality and Knowledge, 125–134, 2007. http://dl.acm.org/citation.cfm?id=1324268. While that paper doesn't mention the FBC by name, it does all the heavy proof needed to see which of the systems on this page comply; the missing steps are trivial. Homunq (࿓) 18:27, 21 April 2014 (UTC)


 * Oppose/comment. Only here (briefly) for the RFC. I doubt I have much to add. Certainly I agree that voting tactics are a non-trivial topic, worthy of appropriate inclusion in WP, and relevant to voting systems. However, as I skim the topic of Voting system, it seems that there is such an independent article already, and no move to unite the two articles. Given this, it seems to me that there is no point to a serious treatment of any form of tactical voting here and that where there is reason to mention any forms of tactical voting in context (which there well could be) they should be referred to by links to the Tactical voting article. It is not clear to me why this should be a topic of dissension. But if I am being obtuse, I can only wish you all the joy of the fray. JonRichfield (talk) 18:45, 5 May 2014 (UTC)

Split out criteria and table to separate article and transclude?
I'm not going to do this right now, as it would probably be disruptive to the above RfC. However, as soon as that's done with, I think that the criteria and table should be put in their own separate article, transcluded here. That would simplify and separate the discussion; after all, there have already been attempts to effectively give the table its own talk page, but I think that without splitting it out as a sub-article, such attempts are doomed to fail. Homunq (࿓) 16:10, 20 April 2014 (UTC)


 * Now that the RFC is done, what do people think of this idea? I don't want to do it without some discussion first. Homunq (࿓) 14:59, 15 May 2014 (UTC)

Can we get a source other than “Cardinal Voting: The Way to Escape the Social Choice Impossibility”?
Below are the problems I see with Cardinal Voting: The Way to Escape the Social Choice Impossibility by Sergei Vasiljev:

1. Assumes voter preferences A:87, B56, C:13 on an approval ballot AB means voter’s cardinal and ordinal preferences are A=B (contradiction).

2. Proves non-dictatorship by assuming a voter can approve candidate A to offset the dictator who approves B, but contradicts by assuming same voter cannot approve A in order to satisfy IIA.

Is there a consensus yet among theorists and how can we know to use this article? Is see no affiliations. Saying all cardinal methods pass Arrow's Impossibility seems like a fantastical claim, given that voters having cardinal weights only adds information to the voter model but does not subtract ordinal information from the voter model.

Filingpro (talk) 09:04, 27 October 2013 (UTC)


 * They all "pass" because Arrow's Impossibility theorem applies to preferential voting systems only. It doesn't mean however that cardinal voting should be preferred over preferential voting. Claiming that is like claiming that dictatorship is to be preferred over non-dictatorship. Dictatorship is the easiest way to satisfy IIA. Wat 20 01:32, 13 November 2013 (UTC)
 * Cardinal models that are dictatorial don't pass Arrow's Impossibility. The point is Vasiljev claims a cardinal model for which all Arrow's criterion are in fact interpreted, applied and pass, but the logic seems to me to have errors. This is why I am asking editors if we can find a better source or if there is any consensus yet in the field. I do not understand the idea that cardinal voting models are not preferential when a voter's cardinal preferences A:87, B:56, C:13 implies A>B>C axiomatically. It seems logically they can not be exempt from preferential theorems.
 * I see your point and we agree that in order to pass IIA we must adopt a dictatorial model for cardinals, but that doesn't make sense to me to adopt such a model, for say, approval voting, which is not inherently dictatorial. Approval voting only stipulates to vote for any number of candidates. Only by imposing a particular cutoff regime we make it dictatorial. That's why, I am quite confident, Approval voting fails IIA. Approval voting is a preferential system because the voter may express or disguise preferences arbitrarily.
 * Filingpro (talk) 01:03, 15 November 2013 (UTC)
 * The reason (most) cardinal systems can be said to "pass" all of Arrow's "incompatible" criteria is that in such systems, erasing a candidate from a ballot does not affect the others. However, doing so can leave an obviously-unstrategic ballot; for instance, an approval ballot that approves all or none, or a score ballot that rates the only two options at 36 and 37 instead of 0 and 100. Arrow's theorem does not account for the possibility of such "strategically incompetent"/"non-admissible"/"watered-down" ballots; it assumes that A>B is always just A>B, not A>>>B or A(>)B.
 * As soon as you take (competent) strategy into account, you are back in the realm of irresolvable dilemmas (as Gibbard and Satterthwaite proved). Homunq (࿓) 20:54, 20 March 2014 (UTC)
 * If voters cannot decide their approval cutoff based on the alternatives in the election, the voting system is dictatorial. Because Approval voting is not dictatorial (i.e. voters may vote for at least their most preferred candidate), Approval voting fails IIA. No voting system passes all of Arrow's criterion as long as you assume voters have preferences.
 * Arrow's theorem did not fail to account for cardinal preference because he proves a stronger theorem, for ordinal preferences.
 * My original question is whether we are basing the IIA passage on Vasiljev's publication which I am concerned is not a recognized source, and for which I have pointed out the errors (please see my point #2 in the original posting). Arrow won the Nobel Prize and is peer reviewed.
 * Filingpro (talk) 23:15, 19 May 2014 (UTC)

Counterexample To Qualification Of Majority Judgement Failure Of Later-No-Harm
The article asserts: "Majority judgment meets a related, weaker criterion: ranking an additional candidate below the median grade (rather than your own grade) of your favorite candidate, cannot harm your favorite."

Counterexample: 6 voters

1: A=”Excellent”, B=”Fair”, C=”Poor” 1: A=”Excellent”, B=”Poor”, C=”Poor” 2: B=”Excellent”, A=”Fair”, C=”Poor” 2: C=”Excellent”, A=”Good”, B=”Good”

Candidate Ratings A: “Excellent, Excellent, Good, Good, Fair, Fair” B: “Excellent, Excellent, Good, Good, Fair, Poor” C: “Excellent, Excellent, Poor, Poor, Poor, Poor”

NOTE: B strong supporters harm themselves by rating A “Fair”, despite being below B’s median grade “Good”.

I will wait for anyone to point out an error with the counterexample before updating the article appropriately. Filingpro (talk) 08:56, 21 May 2014 (UTC)


 * 1. This counterexample is not well-stated, but you are correct that it points to a true counterexample for Majority Judgment. Later-no-harm is an individual, not a collective, criterion, so you must change your ballots so that 1 of the B voters has already strategized in order to get a counterexample for the other B voter. In general, counterexamples only exist when (among other requirements) the pivotal voter is the only one to give the relevant candidate a certain grade. In a large election, that is asymptotically never.
 * 2. The footnote itself is based on Balinski, M., and R. Laraki. “A Theory of Measuring, Electing, and Ranking.” Proceedings of the National Academy of Sciences 104, no. 21 (2007): 8720, in which they show that the stated qualification applies to "order functions" including the median, but do not address the Majority Judgment tiebreaker procedure. Since I myself usually focus on median systems for which the footnote would be true¹, I didn't realize that Balinski and Laraki were pulling a bit of a fast one here.
 * 3. I still believe that something qualification of the current footnote is appropriate and supported by sources. I'll try to edit it so it does not misstate the facts.
 * ¹Such procedures include: Graduated Majority Judgment, in which median ratings are linearly interpolated between the grade transitions; Majority Approval Voting, which breaks ties by number of above-median votes; ER-Bucklin, which breaks them by number of median-or-above votes; Majority Choice Approval, another name for ER-Bucklin with only 3 grades; Continuous Median, in which grades are set on a continuous scale using a slider, and so no tiebreaker is needed; and Bucklin-Condorcet, in which a Condorcet matrix is built from the top grades down until some candidate has a majority. Of all of these numerous median procedures, MJ is the only one where a counterexample like yours works. I realize that all of that is WP:OR; but then, so is your counterexample. I think it's OK to discuss original findings here on talk (for instance, I'm grateful for your counterexample), as long as they're not the basis for claims on the page. Homunq (࿓) 12:21, 21 May 2014 (UTC)
 * OK. Good but the footnote is confusing "Median systems like MJ meet a weaker criterion..." this means or at minimum implies MJ meets the weaker criterion. Can we fix that? I'll put in a hack for now but please smooth it out. Thanks.
 * Filingpro (talk) 07:48, 22 May 2014 (UTC)

RE: Edits Qualifying IIA Compliance for Rating Methods
Re: Homunq edit to the posting

I object to the word "voluntary" when the voter abstains in an election due to an immovable approval cutoff being imposed on the voter by the mathematical model.
 * 1

The reason is that the voter does not choose the alternatives in the election, and so does not choose whether they will vote in any particular election (i.e. they have no control over whether they vote or not). Whether or not they can vote is determined by factors completely external to the choices available in the election.

When we stipulate a model of the voter using an absolute cutoff, we have invented a creature for which we can not ascertain the motives unless we artificially ascribe them - i.e. in my posting I am not describing an "expressive" voter who "volunteers" to abstain. I am merely saying if Approval passes IIA, the cutoff must be absolute, and if so we can only conclude that a voter must necessarily abstain in certain elections, not whether the voter does so voluntarily.

Likewise I object to the edit re: voters "foregoing their voting power". If the absolute cutoff is computed prior to knowing any available candidates, the voter has no power to exercise or forego in the election.

The real issue is that Approval voting fails IIA, both in the real world of course, and also it fails mathematically because any non-dictatorial cutoff model fails. It is only when we choose to adopt a dictatorial cutoff model that we can claim it passes IIA, which I don't see why this model would be chosen. We would have to add "Non-dictatorship" criterion to the table, showing Approval's failure. But Approval voting is not a dictatorship. It merely says "Vote for any number of available alternatives."

I object to broadening the final statement, in this context, to saying all voting systems fail IIA. We are not talking about all voting systems here. We are talking about rating methods and whether they pass or fail.
 * 2

I do support adding the comment regarding how all deterministic, candidate-neutral methods fail IIA, but this valuable comment belongs in a section of the article that addresses IIA in general. If no non-dictatorial systems pass IIA, then IIA compliance under the rating methods shouldn't say 'Yes'. To me, this looks like an obvious error on Wikipedia. We know this because of Arrow's renowned work for any voter preference aggregation system.

I will allow for comments before reverting the edit by Homunq Filingpro (talk) 18:21, 20 May 2014 (UTC)


 * We can either fight over this, or find a compromise everyone can live with. If it's the former, the truth doesn't matter; only what we can WP:CITE — and citations seem as absent from your argument above as from your edits to the page. In other words, if either one of us tries to WP:OWN this or starts an WP:EDITWAR, I think we'll both lose.
 * Since I prefer it when the truth matters, I'd rather take things in the spirit of compromise. I didn't revert your edit, but rather tried to craft a compromise which would separate the baby from the bathwater. I'd appreciate it if you'd do the same with what I proposed.
 * So, to address your specific arguments:
 * 1: I'm not attached to the word "voluntary", but I don't think the word "dictatorial" has any place here either. Approval can pass IIA without any external force imposing a universal cutoff; it works just as well if every voter has an internally-motivated, voluntary cutoff. Both of those are mere words around the idea that the cutoff does not vary depending on the candidates available; let's just describe that as neutrally as possible.
 * 2: As Tonto said, "what do you mean, 'we'?" If an article about human life span said that "no Scot is immortal", it would be appropriate to edit that to say "no human is immortal"; needlessly over-qualifying a statement in this way is misleading.
 * I look forward to seeing your proposed text, and to working to find something that we, and everyone else who cares, can live with.
 * Homunq (࿓) 01:30, 21 May 2014 (UTC)
 * I removed "Dictatorship"
 * re (1) Let me clarify, the voter cutoff may be chosen by the voter, but when it must be computed independently of knowing the available candidates, then the voter's ability to vote in an election with two arbitrary candidates is random, the voter has no control over abstention.
 * re (2) I strongly disagree. The rhetorical method you employ is genius, [1. Approval passes IIA under certain assumptions] then, after softening these assumptions, you imply but do not say explicitly [2. By the way, these assumptions are reasonable because if we lift these assumptions then any method would fail IIA]. In other words, you turn it into a proof ad absurdum, that lifting these assumptions doesn't allow us to distinguish methods that pass or fail IIA (when the truth is, obviously, none of them pass). This obfuscates, in a brilliant way, the truth. Here is the truth: Passage of IIA by rating methods requires extreme, unrealistic assumptions. When they are lifted, they all fail IIA.
 * re edit war - I am not concerned - I take a long term view and I'm not likely to spend much time on this as I am working on a publication - feel free to improve my edit. I totally accept that we may have different and strong opinions. No problem. Thanks for working on the article. I have learned a lot by working on Wikipedia.
 * Filingpro (talk) 10:01, 21 May 2014 (UTC)
 * I edited the page again. I think we're coming closer to something we can agree on. As long as we don't do out-and-out reverts, I'm happy doing this through page edits alone; I think we understand each others' positions well enough to skip the talk page on this issue from here on. Homunq (࿓) 12:38, 21 May 2014 (UTC)
 * Excellent. I like "For this to hold, in some elections, some voters must use less than their full voting power or even effectively abstain, despite having meaningful preferences among the available alternatives."
 * One important final objection: The reader has a right to know the minimum to fail IIA, otherwise we obscure the truth about the conditions of its failure.
 * Since we've gone back and forth on the last sentence, how about this: if you can provide reasoning or proof that failure IIA requires some voters to both put their favorite highest and and least favorite lowest, then we can put that in the article.
 * Meanwhile here is an example where a weaker minimum condition causes the IIA failure (actually its not a condition its a lack of restriction)...
 * Example - In Majority Judgement, abstaining from rating a candidate gives them the lowest rating. Assume ratings: Excellent, Good, Fair, Poor
 * 3 voters prefer A>B>C vote A=Fair, B=Poor, C=Poor
 * 2 voters prefer B>C>A vote B=Fair, C=Poor, A=Poor
 * 2 voters prefer C>B>A vote C=Fair, B=Poor, A=Poor
 * BC coalition vote split, A wins, Condorcet winner tie BC lost
 * Remove C from the race. Hold another election. C supporters need only give any positive rating, e.g. "Fair", to their favorite candidate in the race (B), to make B the winner, failing IIA (changing the winner from A to B by removing and adding spoiler C from/to the race).
 * Rather than back and forth edit, here is what I propose, for the last sentence:
 * "If, on the contrary, it is assumed at least possible that any voter might give a minimally positive rating to their favorite candidate, these systems fail IIA."
 * Thanks if you can come up with something better.
 * Filingpro (talk) 18:53, 22 May 2014 (UTC)
 * Probably a bad idea to spend too much effort being precise in our OR. I'll make another simpler try. Homunq (࿓) 19:38, 22 May 2014 (UTC)

Majority Judgment Qualified Failure Condorcet Loser
Article reads:"Majority judgment will not elect a majority Condorcet loser, but can elect a Condorcet loser if some voters are indifferent."


 * 1) 1 Here is an example of MJ failure of Condorcet Loser, but I don't see any "indifferent" voters:

https://en.wikipedia.org/wiki/Condorcet_loser_criterion#Majority_Judgment


 * 1) 2 What is a "majority Condorcet loser"? Where is that term defined? What is the proof, or reasoning? Is this WP:OR? How do we know this is relevant or correct?

To be clear, I am open to a qualified failure awarded, but I don't understand the explanation given. I am concerned because it seems clear the statement includes false claims. Thanks for any clarification. Filingpro (talk) 02:25, 23 May 2014 (UTC)


 * You are correct. Homunq (࿓) 02:35, 23 May 2014 (UTC)

Approval & Majority Judgement Qualified Failure of Majority Criterion
Article: "Approval only passes the majority criterion if the majority approve of only one candidate. Though this is strategically rational of them if they know each other's preferences, it may not be the obvious strategy if they do not."

Issue #1 1st sentence seems to be false

Here is an example where Approval passes majority without the majority approving only only candidate...

40 voters A>B>C approve A 30 voters A>B>C approve AB 25 voters B>A>C approve BA 5 voters B>A>C approve B

Majority preferred candidate A wins 95 approvals to B 60 approvals

Issue #2 "Though this is strategically rational of them..." I assume this sentence is the justification for its qualified failure? That is, as long as voters are strategically rational, it passes Majority?

The problem with this is what consistent standard do we apply for qualified failures and passage of criterion?

For example, Borda passes Majority if the Majority strategically votes only their preferred candidate.

Concerned also the same for Majority Judgement.

I see a contradiction because Approval and MJ fail Pareto Efficiency (if every voter prefers A, the voting system must elect A), yet they receive a qualified failure for Majority criterion. Meanwhile, Borda satisfies Pareto Efficiency, but isn't granted any privilege with respect to the Majority criterion. Filingpro (talk) 21:43, 22 May 2014 (UTC)
 * We've had this discussion before. Approval passes the majority criterion, period. That is, if a majority vote candidate A above all other candidates, A wins. This is not true of Borda.
 * Obviously, it wouldn't be appropriate to give it a green, as we both understand, because this technical meaning of the majority criterion is not what most people, including you, think it means. Most people would think that preferences that are in your head but not on the ballot should count for something. But that's not how criteria are defined, so it would be equally inappropriate to completely give it a red.
 * The current wording is a compromise. If you think you can do better, go for it. Homunq (࿓) 22:21, 22 May 2014 (UTC)


 * Criterion are not defined by the ballot (ballots change), they're defined by the framework (i.e. mathematical model) used by the author. As editors, we read the terms and definitions of the author. For example, Woodall defines a voter ordinal model. He defines the literal 'ballot' to be a listing of voter's ordinal preferences, as might be cast in an election. You can not remove the voter's ordinal preferences from the model.
 * Since the statement in the article is false I will remove it.
 * Since the sentence about strategic voting doesn't seem to be the reason for the qualified failure (because then Borda will be qualified failure) I will remove it too.
 * Filingpro (talk) 01:19, 23 May 2014 (UTC)
 * I think you are just wrong here, but it's not worth fighting about. The statement is not vital to the article. Homunq (࿓) 02:36, 23 May 2014 (UTC)

Interesting reference
has a table like the one here... we should certainly use it as a reference when possible. Homunq (࿓) 12:59, 5 June 2014 (UTC)

Majority "Rated" Consistency And Justification WP:OR Problem
@Homunq (or others)

Re: "rated" definition of Majority "candidate A voted above all others" vs. common definition "candidate A preferred by voters"

Why must we stipulate the strict voting above on the ballot, rather than the voters preferences?

None of the other global criteria (Condorect, Mutual Majority) do we do this. Of course, as we should, we consider the voters preference because they are the input to the voting system, and we wish to compare how voting systems respond differently to the same input.

Why is Majority different?

How can we justify applying the strict ballot markings (0, 1) as the basis for interpretation of preference for Majority, but not Condorcet and Mutual Majority? (Approval voting) Filingpro (talk) 16:32, 27 July 2014 (UTC)

Inconsistent Reasons For Majority Judgment (MJ) Qualified Compliance
1 MJ receives qualified failure of Participation because voters may strategically change their ratings, based on information on how others will vote.

2 MJ receives qualified failure of Consistency assuming voters use absolute ratings (not strategic).

3 MJ receives qualified passage of IIA assuming voters use absolute ratings.

Which is it, strategic or absolute?

Are absolute ratings a viable model? Must a voter use the same ratings they mark in a closed party primary election, in the runoff election?

"Semi-honest" strategy requires the voter to lie strategically, giving candidates relative ratings they would not otherwise, i.e ratings they do not deserve. How does the voter know to do that? How does the voter know they will not harm themselves? Please see problem discussed above "What is our standard for qualified failure?"

Filingpro (talk) 07:42, 27 August 2014 (UTC)

Electoral Systems Maps lacking a color code legend, date or source
Two colored "electoral system maps" of the world appear in the article at section 5.5 "Post-1980 developments". There appears to be no legend (A key to the symbols and color codes on a map, chart, etc.) to accompany the maps. In addition there is no date for the maps which is important as jurisdictions do change voting systems. Finally there is no apparent source or authority for the maps. In my opinion the maps are of little value without these 3 features but I have not deleted the maps in the hope that another user has the missing information to include in the article. Lanyon (talk) 07:46, 19 May 2015 (UTC)

Runoff voting and Condorcet loser
On 13 May 2014, the Condorcet loser box for runoff voting was changed from "yes" to "no". I am surprised that such a blatant error wasn't corrected for more than a year. Markus Schulze 10:56, 17 September 2015 (UTC)

What is our standard for qualified failure of Condorcet?
All voting systems pass Condorcet when voters have perfect information and vote strategically, because those who support the Condorcet Winner bullet vote for the Condorcet winner. (This has been discussed, illustrated by examples, and acknowledged in a more detailed discussion in the Archive).

Q: If they all pass Condorcet, including, for example IRV, why are there seemingly arbitrary qualified failures? What is our standard?

If we have a standard, I think we might consider that a method should not receive the benefit of a qualified failure when it fails the criterion under common circumstances under which another method without qualified failure (i.e. full failure) passes the criterion.

For example, current MJ is granted qualified failure and IRV full failure. But here is a common example of the "center squeeze" problem in voting that IRV handles elegantly, but Majority Judgement loses the Condorcet winner.

L = extreme left candidate CL = center left candidate R = right candidate

Ratings: "Excellent, Good, Fair, Poor"

40 left voters L>CL>R, vote L=Excellent, CL=Good, R=Poor 32 center left voters CL>L>R, vote CL=Excellent, L=Good, R=Poor 28 right voters R>CL>L, vote R=Excellent, CL=Fair, L=Poor

L tallies:40 Excellent, 32 Good, 28 Poor CL tallies: 32 Excellent, 40 Good, 28 Fair R tallies: 28 Excellent, 72 Poor

Majority Judgement count: Tie between L and CL with median "Good". Remove 21 "Good" ratings. L is closest-to-median winner "Excellent".

IRV elects the strong Condorcet winner CL. Majority Judgement elects extremist left candidate L. Filingpro (talk) 22:54, 22 May 2014 (UTC)


 * The standard is, is there always a possible strong semi-honest equilibrium for a CW. This does not hold for IRV or Borda, but does hold for any system which can reduce to approval (including Range, MJ, Approval... and also things like Schulze and RP, though the latter pass the CC without needing this anyway.)
 * As for your example, it's even easier to find examples where honest IRV fails Condorcet and MJ doesn't. (Also, your example is a chicken dilemma, but not a center squeeze. IRV fails center squeeze; it means that the centrist has fewer partisans than either winger.) Homunq (࿓) 02:43, 23 May 2014 (UTC)
 * MJ can also fail "center squeeze" scenarios- e.g. change 4 center left (CL) voters above to right (R).Filingpro (talk) 21:17, 30 August 2014 (UTC)
 * I don't think its a Chicken Dilemma because L and CL can not cause R to win by withholding their support for each other. I believe its a straightforward case of loss of the Condorcet winner with sincere voting. Filingpro (talk) 21:17, 30 August 2014 (UTC)
 * sure understood center squeeze used differently that I do here - no problem
 * For our standard, why does the equilibrium have to be "semi-honest"?
 * Note: R supporters (above) cannot just promote CL to "Good" to improve the outcome, they have to lie and give CL "Excellent", the same rating as their favorite (R). The ballot markings could mean voter ordinal preferences R>CL or CL<R with 50% probability each. We wouldn't know the ordinal preferences. I can come up with cases, in IRV, of correcting a Condorcet winner by fewer ordinal reversals, hence a more honest profile of the voters sincere preferences.
 * Filingpro (talk) 03:37, 23 May 2014 (UTC)
 * The equilibrium I'm referring to is, in the case of your example:
 * 40 left voters vote L=Excellent, CL=Good, R=Poor
 * 21 (or more) center left vote CL=Excellent, L=Poor, R=Poor
 * 11 (or fewer) center left voters vote CL=Excellent, L=Good, R=Poor
 * 28 right voters R>CL>L, vote R=Excellent, CL=Fair, L=Poor
 * Note that even if only 11 of the CL voters truncate L, this result is already nearly assured; though it's not, strictly speaking, a strong Nash equilibrium until there are 21 such votes.
 * Homunq (࿓) 18:28, 23 May 2014 (UTC)
 * Sure I see that too, but it requires voters to lie about their judgments, and more importantly, to do so strategically.
 * If the qualified failure is awarded based on strategy leading to compliance, IRV leads to compliance under strategy.
 * Why does "semi-honest" matter when only compliance matters? In other words, when voters are consciously employing strategy, why does it matter how they do it?
 * What if Majority Judgment has three ratings "Good, OK, Bad"? Everyone votes again in the election above. Where is the equilibrium? Center left must lie, betraying their coalition, giving their second preferred L candidate the same rating "Bad" as the R candidate. That's a lie, because they are not equally preferred. There is nothing "semi-honest" about that. Statistically speaking, with regard to voter ordinals, 50% could be L>R, and 50% could be R>L! If you assume the same ratings "Bad" to be equal voter preferences (a false assumption), then that's a lie because voters actually prefer L>R, not L=R that's a lie.
 * Filingpro (talk) 19:08, 24 May 2014 (UTC)
 * Arguing your position, I might try to claim that its easier for voters to employ strategy with rating ballots than IRV. But the problem with that assertion is two fold. One, is its WP:OR opinion, and I'm not sure we have enough evidence to back up that claim. Two, consider the example above. How can center left voters be sure that rating their second preferred candidate L "Bad" will not elect their least preferred candidate "R"? Not so easy. This strategy requires a knowledge of how ballots are counted and how others will vote. We could just as well argue, that under a different IRV "center squeeze" election, a right voter knowing their least favorite candidate will win 'L' instinctively vote 'CL>R' as a defensive move, to see if they can get a better outcome in the Nash equilibrium. We would also need evidence to show voters who have perfect information under Nash strategy can not instinctively understand early elimination, (or whether or not they even have to understand it to know to try promoting their second favorite in the Nash game). This all seems way beyond what we can put in the article.
 * Filingpro (talk) 19:34, 24 May 2014 (UTC)
 * Perhaps the distinction you would like to make is now covered by the inclusion of FBC (I prefer Kevin Venzke's "Sincere Favorite"), and is now redundant. (Actually I think would prefer the name "First Choice First" or "First No Harm").
 * Filingpro (talk) 19:42, 24 May 2014 (UTC)

FINAL COMMENTS BEFORE CHANGE (WILL WAIT 2 WEEKS) PROPOSAL TO REMOVE CONDITIONAL FAILURE OF CONDORCET BY RATING METHODS

See also my comments above which have not been rebutted.

Consider 10 approval voters with three alternatives A, B, C, with cardinal preferences from 0 to 100, and an approval cut off of 50:

3 A (cardinal preferences A:100, B:40, C:0, approve only A) 3 B (cardinal preferences B:100, A:45, C:0, approve only B) 4 C (cardinal preferences C:100, B:10, A:0, approve only C)

C, the Condorcet loser is elected. B is the Condorcet winner.

The proposed standard (for conditional failure) is, is there always a possible strong semi-honest equilibrium for a CW?

Semi-Honest Oxymoron In order for Condorcet Winner B to win, “A Voters” (or "C Voters") must know to approve B. But B is below their cutoff. In fact B is closer in rating to their most opposed candidate! Approving B is a lie, or at best insincere. (Examples I gave above with multi-level ratings and MJ are perhaps even more dishonest, because the voter has discrete levels to choose from yet disregards them). Burr Dilemma How do B voters know not to approve A (5 points below their threshold), to at least change the result from C to A, a better outcome? Can they depend on A voters to pull through and Approve B (10 points below their threshold)?

How do A voters know to vote for B, because B voters like A voters more than A voters like B voters. Will B voters compromise?

This appears to be a potentially unstable situation, where either A voters or B voters might benefit from a coalition betrayal as they face a game of chicken.

The Burr Dilemma has been well documented by professor Jack Nagel and cited.

I object to the use of Nash-Equilibrium (for use in the qualified failure) as it implies voters continuously adjusting their ballots with feedback until an optimal solution is obtained. I object for two reasons: (1) This is not what happens in a real election (i.e. continual adjustment) - in practice voters get to vote once and can not modify their vote once cast, and (2) Under such a feedback process IRV also reaches equilibrium (perhaps more efficiently and without the Burr Dilemma). Under continual adjustment, there is no evidence that it would not be equally obvious to an IRV voter to compromise vote for the Condorcet Winner if they could achieve a better outcome. There is no evidence that rating ballot voters would have any easier time knowing to change their intended sincere ratings. Both cases involve the act of raising their second choice to try to get a better outcome than their last choice, and knowing when to do so.

Thus I see and have heard no justification for awarding qualified failure to rating methods and not IRV.

Lastly, I have pointed out common voter scenarios where Approval, MJ etc, fail Condorcet (cases where there are simple vote splits) which IRV handles elegantly. Yet somehow these rating methods are elevated in the compliance table.

I propose removing these qualified failures. Filingpro (talk) 03:31, 6 May 2015 (UTC)

Removed qualified failures of Approval, MJ, Range. Also removed qualified failure of MJ for Participation based on exact same reasoning argued above.Filingpro (talk) 07:41, 27 September 2015 (UTC)

Compliance table - exhaustive balloting, block voting, preferential block voting, and STV
I don't have the required knowledge nor references to make a contribution on this, but I noticed the compliance table is missing a number of common systems, specifically:
 * Exhaustive ballot
 * Plurality-at-large voting
 * Preferential block voting
 * STV

While none of the systems currently listed have identical attributes and while these systems may share the same attributes as another on the table, their popularity is great enough that they should not be omitted. Is someone with more knowledge and references able to add those to the table? 24.246.28.27 (talk) 00:29, 6 December 2015 (UTC)


 * The compliance table applies to single-winner systems only. Markus Schulze 10:50, 6 December 2015 (UTC)


 * D'oh. I missed that statement. What are your thoughts on a similar table for multi-winner systems? Some criteria don't seem to apply (although I could be wrong) but some criteria do seem to apply (although I could be wrong about that, too). 24.246.28.27 (talk) 03:00, 11 December 2015 (UTC)


 * Criteria for multi-winner systems are not well-defined, so a similar table for multi-winner methods would be even more controversial than this single-winner table. VoteFair (talk) 05:41, 13 December 2015 (UTC)


 * However, a compliance table for multi-winner methods, even partial and discussing controversial points, would be very useful. For example STV is accused not to satisfy many criteria, like monotonicity, consistency, participation, indipendence of irrilevant alternatives, vulnerability to various strategic voting. It seems that Schulze-STV and Satisfaction Approval Voting don't suffer of many of these problems, though they respect proportionality (or they "waste" less votes like STV is aimed to). Could you help for such a new compliance table? Armando Pitocco (talk) 07:33, 16 July 2016 (UTC)


 * Yes such a table would be useful. But Wikipedia is an encyclopedia so the facts conveyed in the table must be established facts, not opinions. Such facts are not available.


 * Note that "wasted votes" is not defined, even for single-winner methods. Also note that if "proportionality" regarding (ballot-supplied) political-party support was defined as a criterion, every method would fail because no method can achieve full proportionality, and there is no established way to assign a number that indicates a "distance" from full proportionality.


 * Again, the yes/no information for the criteria presented in the current table are well-defined for single-winner methods (with a few small exceptions), but not for multi-winner methods. VoteFair (talk) 20:01, 17 July 2016 (UTC)


 * I think that some criteria could be defined also in multi-winner methods. For example traditional STV (multi-winner) is non-monotonic, at least is what I know and read in a paper linked in the italian page. If we can't say yes/no for some methods we can leave a "?" and put an explaining note. About proportional representantion wikipedia mention PR-list and STV, so it is an existing "property" (maybe not strictly a criterion). About wasted votes, maybe we can find a consensus to explicit this concept, since it is an other commonly recognized property. For example a PR-list method without threshold waste less votes than one with a 5% threshold.


 * What I'm saying is that some properties exist, and where possible I think we should describe them. Where not we could explain why. Maybe we can't draw a compliance table, since it isn't so clear the situation? We can write a paragraph about methods comparisons.Armando Pitocco (talk) 20:44, 17 July 2016 (UTC)


 * I looked at Single transferable vote and it says STV fails monotonicity.
 * Another consideration is that there are many proportional methods, but the only "notable" ones are the ones actually used for governmental elections, and they are seriously flawed, mostly because they use single-mark ballots (or equivalent). In contrast, the better methods have not yet been used, and only a few of them have been written about in academic journals, so those are not yet notable and carefully analyzed.  The result is that your suggested table would contain only a few rows (and only a couple of columns).
 * I agree that before jumping to a table it is necessary to create content in text form. There is not enough information to justify creating a table about proportional methods.
 * I share your desire to bring more attention to better proportional methods. After all, I've created a semi-proportional system called VoteFair ranking .  (In my opinion methods that are "fully proportional" according to political party lead to fragmentation into too many parties to be handled using the usual coalition approach.)  But to justify adding content on Wikipedia requires that first the methods must be better studied and analyzed, then written about in academic journals, and then the details can be summarized here. VoteFair (talk) 20:09, 20 July 2016 (UTC)
 * The methods we have talked about are already in Wikipedia singularly, so there is no problem in writing a paragraph to explain differences, according to what actual researches can say. In explaining the differences we can refer to some of single-winner criteria, where sources are enough. Armando Pitocco (talk) 16:49, 21 July 2016 (UTC)