Talk:Voting system/Archive 1

From Wikipedia, the free encyclopedia

Archive This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Contents

coverage

I know I've been scattering pages all over the place here, but here's my idea. Voting system will be used to discuss the dry, mathematical aspects, election to discuss specifically human elections, probably their history, geography, etc. Each system page will discuss the mathematical aspects of that particular system, as well as examples of it's use (real-world and hypothetical). I'm not sure how much of a Proportional representation to recreate on [Voting system]. --DanKeshet

More about methods, I think that someone who knows should perhaps add notes about the adducted reasons for every system, in order to describe why it is respectively supposed that each system would be a better instrument to represent electors rather than the other ones (thus expanding what is now in the considerations list). At least a concise list of alleged pro and contra, so that we don't stop at a technical point.
I.e.: I confusely remember Kelsen's notes about proportional system, where he describes why he thinks this should ensure the better popular representation and respect of minorities etc., while others instead criticise this because... Others propose let's say a "majority prize" stressing this should ensure a more effectively productive parliamentar or governmental life, while oppositors believe that... and so on.
I know that this would risk to become and endless tale, and also that (due to single parties' positions on the matter) it would always keep us on the border of a non-NPOV, but a bit of detail (scientifically reported) would render a more complete idea.
Could this eventually belong to a separate page?

context

Voting systems are usually discussed in the context of electoral reform or grassroots democracy, or as an alternative to consensus decision making, but the tradeoffs in choosing a voting system are so complex that there is no easy summary of how these various debates interact with system characteristics.

Removed because the first sentence is not necessarily true; there are many environments in which voting systems are discussed; the second part is discussed already, and better, in the paragraphs below. pde((?)).

coalitions

As it stands, we have this text regarding coalitions:

These parties may in turn be aligned with other parties, to form coalitions, which can play roles beyond those played by the party.

But I'm wondering, are coalitions ever actually a formal part of the voting system, and not just political or parliamentary entitites which make agreements on candidates and/or lists to submit to a voting system (in which case, they're a bit off-topic)?

--Pde 04:17 8 Jul 2003 (UTC)

In the Australian(?) system, you can fill in a set of preferences, or you can alternatively just check the "Green" checkbox and have the Green party fill in your preferences for you. In such a system coalitions have a considerable impact.
Also, some voting systems encourage or discourage coalition government (eg proportional representation tends to create coalitions, first past the post tends to create a two party systeam). So there are links that way too. Martin
Yes, I meant those words exactly as you understood them. See, for example, http://www.kypros.org/Elections/electoral.html In that system, you can vote for single-party lists or multiple-party coalitions, and there are different thresholds for single-party lists and for coalitions, and even for coalitions based on their size. Cyprus is not unique in this regard. DanKeshet 14:50 8 Jul 2003 (UTC)

A grumble about a change that I'm not bold enough to undo myself: disapproval voting is a valid voting topic but is not a "system" in the same sense as the others listed here.

-- pm67nz

want to add a section for how the absentee ballot is handled in various states

Hi,

I'm new to the wiki as an author and I would like to contribute a section on the votes are actually sorted and tabulated in various states. Specificly, I wanted to add a section on how absentee ballots are handled and then tabulate which states allow for absentee ballots. The current Voting System page doesn't allow for this very well without an addition to the table of contents. My thought on this is to add a voting methods section under the aspects of voting systems and have 3 sections:

Voting methods
Absentee voting:
Internet Voting:
DRE voting:
Paper ballot:

Then I will want to expand on the sub headings and get others who are well versed to do the same.

Thoughts? --Ceremona 19:56, 28 Feb 2004 (UTC)

classification scheme

I do not understand your changes to the classification scheme, Tom:

  • In what sense is the borda count an "approval" scheme.
  • All single-winner systems by definition are "majoritarian" (as opposed to proportional).
  • The bloc vote is not semi-proportional, it is majoritarian.

Please explain these changes.

DanKeshet 18:42, May 7, 2004 (UTC)

Dan,

Proportional Representation says Electoral systems that do not result in proportional representation are known as majoritarian systems.

The term "majoritarian" would seem to me to only apply to multiwinner elections. Maybe this means majoritarian systems pass this test: "If a majority of voters agree-to-agree before the election, they can guarantee to control all the winners."

If so, I accept that this would be true for all reasonable single-vote and approval-vote methods. Still I think differentiating multiple vote methods is important.

Imagine running an election poll by two methods: Plurality and Method X. If a single candidate gains a sincere majority in Plurality, that candidate should win any reasonable election method. Unfortunately that candidate is not guaranteed to win with an Approval or Borda count because that majority is not necesarily aware of its power and they may "overvote" and cause their favorite to lose. At least Bucklin can pass this test since it only allows each voter one vote in the first round.

Borda is an "approval" method because it is not a "One person, one vote" method. Once voters offer more than one vote, there's no definition of majority winner. The only question I have about Borda, is "Can voters truncate their preferences?" The answer doesn't change that it is an approval ballot. It only changes how much Borda looks like simple approval.

Looking at simple approval voting: If I approve of candidates in the order of A,B,C, and dislike D and E, I might approval vote [ABC] counted as: (A=1, B=1, C=1, D=0, E=0); Or approval vote [AB] is counted as (A=1, B=1, C=0, D=0, E=0); Or approval vote [A] counted as (A=1, B=0, C=0, D=0, E=0)

In comparison in the same Borda election, allowing up to 5 rankings, I might rank vote [ABC], counted as: (A=5, B=4, C=3, D=0, E=0); Or rank vote [AB], counted as (A=5, B=4, C=0, D=0, E=0); Or rank vote as [A] counted as (A=5, B=0, C=0, D=0, E=0).

So you can see Borda looks very much like Approval - the number of votes offered increases as more candidates are "approved".

As to the multiwinner elections, I apologize. I was in error. The method "Bloc Voting" is apparently identical to the method I call "Plurality-at-large" which I added. I should remove Plurality-at-large and add it as another name for Bloc Voting.

Tomruen 02:07, 8 May 2004 (UTC)

Regarding bloc voting, thanks for moving that to majoritarian. If anything, it's super-majoritarian because it can produce results even more lopsided than SMDP can.
Regarding single-winner methods: I can understand the distinction you're making, although I think there are better ways to categorize them. More importantly than the quality, though: is this something you have seen in other literature or are these categories idiosyncratic? If it's an accepted system (like majoritarian/semi-proportional/proportional is for multi-winner systems), then let's do it. But if it's just a new system that you made up, I think it's better not to impose our own system of categorizing on these systems. DanKeshet 02:40, May 8, 2004 (UTC)

Hi Dan,

Sorry, I apparently didn't have a watch on this talk page.

I'm not sure what super-majoritarian would be. Bloc-voting allows a devoted 51% to win all the seats. Maybe you mean that you don't even need 51% to get all the seats since there's no majority requirement for winners. I don't know what SMDP is.

About the single-winner method groupings, admittingly the categories I offered may not be any sort of consensus position. I have read about some methods being considered "Approval methods" as methods that offer more than one vote at a time, although some may argue that Condorcet is an approval method as well, while I keep it separate. Whatever classification is used, I really do believe it is useful to separate methods that can be judged "single vote" versus "multiple vote". Single vote methods (Plurality, runoff, IRV, STV, Cumulative Voting) are all methods used in political elections. I don't know any multivote, single-winner election method used in politics. All proportional represenatation systems are (or should be) single-vote methods - every voter gets equal votes to influence the results.

Anyway, I'll do some more research in some books and online and see what I can find. I suppose it might just be useful to point out (elsewhere) which methods are used in various locations, or if they are just theoretical.

--Tomruen 19:34, 17 May 2004 (UTC)

Hi Tom,
RE: bloc voting: SMDP is Single Member District Plurality (first-past-the-post, but emphasizing the system-wide aspect of it). While first-past-the-post is majoritarian in a single district, often because of the variation in districts, it has some aspect of proportionality (e.g. there are 2 districts where D's are in the majority and 1 where R's are in the majority). In bloc voting, all 3 districts become one large district, where D's are in the majority and win all three seats. That is what I meant by super-majoritarian.
RE: the distinction you make about "approval methods": I have never heard it before. Also, when you say that in "single-vote methods" (another term I've never heard before), every voter gets equal votes to influence the results, this is also true in what you're calling "approval" methods. All voters are equal in all of these methods. DanKeshet 21:18, May 17, 2004 (UTC)
Hi Dan,
Terminology is always a problem. When I say "single vote method", I mean the same as saying "One person, one vote" - any system that allows every voter one vote for one choice, (or more generally allows a full vote split among a set of candidates). Plurality obviously follows that logic. Runoffs and IRV also follow it since every voter has one vote each round.
Methods like "Approval", "Borda", and "Bucklin", do not follow the "one person, one vote" ideal. Approval allows multiple full votes. Borda forces multiple votes (unless truncation is allowed). Bucklin applies multiple votes if the first preference votes don't identify a majority candidate. In these methods, if you add the total votes applied, it will exceed the number of voters.
You might even define an "Approval method" as a method where two or more "majority winners" may exist, although the definition depends on how you normalize the count. You can normalize the count by dividing by the maximum votes one person can offer to a single candidate, then a majority support is fairly defined as half as many votes as voters.
In "single vote" methods it is possible to identify a majority winner as a singular candidate being supported by more than 50% of the voters. In multivote methods it is possible more than one candidate will exceed more than 50% of the vote (because the total votes exceeds the number of voters).
In Borda, for example, you might say "Everyone gets equal votes", rank 1,2,3 and apply 3,2,1 points, and so everyone offers 6 points. This can be normalized (by dividing by 3) to 1, 2/3, 1/3 point vote for each ranking, so every can be said to be allowed up to 2 total votes, but no more than 1 vote for any one candidate. But the system disallows putting a full allowable vote on one candidate. If I only have one favorite, I must "bullet vote" (offer 1 ranking, 1 vote). Some people get up to 2 votes while I'm limited to 1 vote.
Similar in Bucklin, if I only have one favorite, and I don't rank deeper than one, and there's no first rank majority winner, other voters can get a second vote to influence the winner, while I am disallowed from putting a second vote on my first favorite.
"Single vote" methods allow runoffs because candidates that are weak because they are not popular. Approval methods (methods allowing more votes than winners) can't use runoffs fairly because they encourage tactical voting - strong candidate supporters can bluff support for weak competition to force elimination of actually stronger competition, and then back down their support later when the competion is weakened. It's very possible for a single-vote runoff winner to lose in an approval-vote runoff.
Similarly for PR in multiwinner elections, if you use a multivote method, there's no good definition of a proportional quota in election. Cumulative voting (CV) and STV both are "one vote" methods because all voters can offer a full vote on one candidate. CV is a single-vote method because you can offer your full vote on one choice so 3-votes is actually normalized to 3-(1/3) votes per person. STV is more clearly a one vote method since it is "one vote" that is automatically fractionally split among multiple choices.
I'm not discounting "approval methods" as unconditionally inferior, but I think the distinction is clear and important to realize as fundamentally different properties.
Every single-seat voting system used in politics to my knowledge is a one vote system. Basically plurality or runoff or IRV. Bucklin and ranked ballots was tried in Minnesota city of Duluth in the early 20th century and rejected by the courts for not being a "one person, one vote" system.
See http://www.fairvotemn.org/resources/publications/Municipal%20Voting%20System%20Reform%20article.pdf


--Tomruen 00:05, 18 May 2004 (UTC)

classification scheme - take 2

regroup method list based on category (single vote, multiple vote, and pairiwise votes), I think this new classification scheme is nonstandard and adds more confusion and bias. Calling a runoff election a "single vote" method is odd. After all, in case of a tie you need a runoff for many other kinds of election. Saying that approval voting is not one person one vote is POV, given the common association of that phrase with basic fairness in democracy. In any given election, most people cast lots more than one vote, but for different candidates in different races, but we don't deprecate that practice. I think the root meaning of the phrase is that person A doesn't get to have their votes given more weight than person A (as was the case in the original constitution with respect to slaves). Talking about used exclusively in political elections would need a lot more evidence to back it up, I think. Suggesting that single vote schemes don't involve tactical voting but multiple vote ones do is clearly incorrect, given the major tactical connundrum that minor party supporters face in the current US system.
Interesting idea, and sorry I missed this discussion until the article changed. But I suggest going back to the previous revision. NealMcB 21:39, 2004 Jun 4 (UTC)
I can and will defend my change as aiding in differentiating methods and reduced confusion. These are fundamental differences. Runoffs are "one person, one vote" because a second vote is allowed only after the original vote is invalidated.
On my claim for "used exclusively in political election", the word exclusive risks being an overstatement, but I know of no exceptions. Of course if you get down to city or other lower elections, anything may be possible in some corners. (Admitting I don't know anywhere that "random ballot" is used.)
According to the book "The International IDEA Handbook of Electoral System Design", 1/97, under "The World of Electoral systems", the single winner methods used by country are:
      • FPTP - USA, Canada, UK, India, Japan, South Korea, Russia ... (30+ other countries)
      • Alternative Vote (IRV) - Australia
      • Two Round - France, Cuba, (~20 other countries)
I think it is deceptive to put "Used" and "Theoretical" methods in the same list as if they were all regarded equally in practice.
If it is judged better for this page to simply list methods, then an alternative location for this information might be a a "Election System History" page that classifies all methods used for practical large scale elections (political or otherwise) in various places and times, and give some background to their local rise (and/or fall). For example, I know Bucklin was used in Minnesota city of Duluth in the past until it was judged as breaking the "one person, one vote" ideal.
Some people may wish to redefine "one person, one vote" to include Approval voting as "one person, one vote PER CANDIDATE", but the standard meaning is "One person, one vote PER ELECTED SEAT". It is the definition that allows a singular majority winner to be identified. It is also the definition that allows all Proportional Representation systems to function for multiple winners.
I accept the classification may not be liked by some interested in giving all theoretical election methods equal footing.
I accept my originally labeling "multiple vote methods" as "approval methods" was nonstandard, but the fundamental difference between allowing an individual voter to support one candidate at a time versus supporting multiple candidates is not debatable. It is irresponsible to not make this clear distinction.
I said "Multiple vote methods give voters more flexibility where tactical voting is less of a concern." This is standard understanding. The more flexible voting is, the more important tactical voting is. This is well accepted even by the advocates of Borda and Approval. Approval itself is a tactical degeneration of Cardinal Ratings.
I will be respectful, but I won't retreat and support completely without more discussion. I would hope that others can improve this change if it needs work.
--Tom Ruen 23:22, Jun 4, 2004 (UTC)
Hi Tom, I agree this section is very confusing. Part of my motivation for lumping stuff in here was to have one place for all the terminology that gets discussed; I think it might make a lot of sense for *this* page to discuss categorizations, with a few popular examples. For example, I can think of at least three orthogonal ways to categorize systems:
        • What the result is (e.g., single-winner, multiple-winner, proportional)
          • How balloting is done (e..g., single vote, cardinal rating, ranking)
          • How it is counted (e.g., pairwise, runoff, etc.)
I think if we can be clear(er) about how we categorizing things, the other discussions might be easier. Then, we have a whole 'other page which is an exhaustive list of systems (perhaps labeled according to categorizations other than the top-level one) along with a note of where (if ever) they're used. That is, keep this page abstract, and put concrete info in that one. Its not perfect, but hopefully it will help us move forward. Drernie 01:49, 5 Jun 2004 (UTC)
Hi Ernie, Thanks for your thoughts.
Certainly "single winner" versus "multiple winner" is the top-level categorical separation and this has already been done (even if there is possible overlap.)
I considered separating by "How balloting is done", but that didn't seem as important since each method can have different types of ballots. Runoff and Instant Runoff use different ballots but same results. Condorcet could have rank ballots OR pairwise ballots, but same results.
"How it is counted" is clear to me as second most fundamental next to "how many winners".
If my regroupings can be improved or wording made with more neutrality, I very much welcome input. I don't want to be unnecessarily controversial, or cause senseless revision/reverting wars, now or over time. I hope there can be some agreement better than an alphabetized listing, even if that's also valuable somewhere. Thanks again!
--Tom Ruen 02:15, Jun 5, 2004 (UTC)
I'm in favour of classification by ballot type. What matters is not the debatable "fundamentalness" of a clasification scheme but its usefulness, particularly to encyclopedia readers. Ballot types are easily and quickly understood. There is also less room for controversy. Tom's classification is very controversial (see gathering storm on electoral methods mailing list). Personaly I disagree both with his interpretation of what ' the "one person, one vote" ideal' means and that the main advantage of 'multiple vote' methods is to 'give voters more flexibility where tactical voting is less of a concern'. The third category, is, by its own admission a messy one, since particular variants of it can fall into either of the first two categories. None of these problems arise with a ballot type based classification.
Dealing with the objections above:
I considered separating by "How balloting is done", but that didn't seem as important since each method can have different types of ballots.
For every method there is a simplest possible ballot, that is the only one that matters.
Runoff and Instant Runoff use different ballots but same results.
Eh? See French presidential election, 2002.
Condorcet could have rank ballots OR pairwise ballots, but same results.
I don't belive "pairwise ballot" is a well defined term. Note above point about simplest ballot - Condorcet is a ranked ballot method.
Pm67nz 06:32, 5 Jun 2004 (UTC)

classification scheme - take 3

I offer a replacement groupings based first on "ballot type" and second on "counting type". Hopefully this is less controverial. These distinction are clear.
The subclass "counting type" is less defendable, especially as listed as "one count" and "multiple counts", but single count methods are worthy to point out since they are simpler and helpful voting strategies will be more transparent in a single round.
--Tom Ruen 08:54, Jun 5, 2004 (UTC)
Someone with only an IP address changed the first catagory to Up/Down so that approval can be moved there. I think approval is a cardinal method, limited values 0,1.
They also added a "ties" subcategory which I think is not clear. For example, they put Bucklin under "ties allowed" which is not true for any definition I've seen. Bucklin doesn't allow a voter two votes in the first round. That blocks the ability of Bucklin to idenify a singular a majority winner in the first round!
I leave it up to others to consider these changes, but I disagree. --Tom Ruen 00:15, Jun 6, 2004 (UTC)

orphaned article: Simple Majority Voting

There's an orphaned article, Simple Majority Voting. Someone might want to rescue the material there, or delete it.

I'll add it back for now under Plurality, until someone tells me the difference. Drernie 22:43, 3 Jun 2004 (UTC)
I would rather see it deleted Pm67nz 06:04, 5 Jun 2004 (UTC)

Single Winner Variations SECTION

Does anyone understand this section? Where the heck is this stuff used? I don't think this belongs on a "list" page like Voting_Systems unless it has a link. --Tom Ruen 05:44, Jun 5, 2004 (UTC)

Single Winner Variations
Automatic Equal Ranking Line Option (AERLO)
A voter may mark a line in his/her ranking, meaning that if no one above that line wins, then that voter wants to promote to 1st place all of him/her above-line candidates and have a recount. (In pairwise-count methods the promotion only takes place if, additionally, there's a circular tie containing above-line and below-line candidates).
Automatic Truncation Line Option (ATLO)
A voter may mark a line in him/her ranking, meaning that if no one above the line wins, then that voter wants to drop from him/her ranking all of him/her below-line candidates and have a recount. (In pairwise-count methods the dropping only takes place if, additionally, there's a circular tie containing above-line and below-line candidates).
Agreed, these are AFAIK recent inventions known only to readers of the electoral methods mailing list, and do not (yet!) belong in a general purpose encyclopedia, well certainly not on a core/intro/list page like this one. Pm67nz 06:35, 5 Jun 2004 (UTC)

theory vs. practice/history

In one of the comments above, Tom suggested that we have a separate page to give an overview of voting systems as they have been used in practice and in history. I think this would make a good subsection of this page. For example: "Around the world, the most popular voting systems are .... and ...., which is especially popular in ...." DanKeshet 19:41, 5 Jun 2004 (UTC)

capital letters

Dear Michael Hardy, you wrote: "Would the authors of articles on voting methods please stop worshipping capital letters with such incredibly fanatical intensity. It makes it hard to get links right."

I consider your recent changes to be vandalism. If you are really so upset about capital letters, then it is sufficient to replace e.g. [[Cloneproof Schwartz Sequential Dropping]] by [[Cloneproof Schwartz Sequential Dropping|cloneproof Schwartz sequential dropping]]. But the way you replaced capital letters destroyed the links completely. -- Markus Schulze 5 Jan 2005

"... is not guaranteed..."

Henrygb: why did you feel that was POV? I thought it was tautological, based on the preceding paragraph. --Baylink 04:48, 25 Feb 2005 (UTC)

Cleaning up

This article is way too long. You cannot give an uninformed person a complete understanding of all kinds of voting systems in one article, but it seems that lots of people have been trying, and adding their own favorite obscure aspects of voting systems.

I've taken out a few things that clearly didn't belong in the article (the line-option stuff that was already discussed here; a duplicate section on criteria signed by Mike Ossipoff; a section on alternate names for Condorcet stuck in the middle of a bulleted list of methods).

The section on criteria, and the table, should be merged into Voting method criteria. There should be a paragraph pointing there, noting that criteria are an important way to rationally compare voting methods when everything else devolves into political ideologies.

I followed the election-methods list for a while. I appreciate the large volume of contributions those people are making, but some of them should keep in mind that things that are hot topics on that list may nevertheless be too obscure for Wikipedia.

RSpeer 19:47, Mar 28, 2005 (UTC)

Remove University of Texas tutorial

I followed the link to the University of Texas tutorial, and found it to be misinformative, and generally lower-quality than existing tutorials in the external links section. Specifically, it called IRV and Approval "positional methods", but the generally accepted definition of a positional method is one where positions on a ballot correspond directly to points.

IRV is at least ranked, which is a term that is confusable with "positional", but I have no idea what inspired them to put Approval there.

Fahrenheit451, I hope you can understand that I'm not reverting your edit out of spite. I would have removed that link if anyone had posted it and I had followed it. That said, it is because your edits to Borda count have made me suspicious that I followed the link.

If any other external links on this page are equally misinformative, or if they are biased or have other flaws, I would advocate removing them too. I simply haven't looked at them all.

RSpeer 17:38, Apr 8, 2005 (UTC)

That's fine on removing that tutorial. Your other criteria for link removal is interesting in that it is contrary to what you are advocating for the electowiki link: "if they are biased or have other flaws.." That is quite broad and would include many, perhaps most links on wikipedia. Perhaps you are operating on a double standard. --Fahrenheit451 02:10, 31 May 2005 (UTC)

New table

I've made a new table of voting system criteria. Some changes I made are:

  • Make the table in Wiki-style, instead of HTML
  • Spell out the method names instead of using incomprehensible abbreviations
  • To accomodate that, split the criteria into two tables (ones that don't assume strategic manipulation and ones that do, though I acknowledge the line is a bit fuzzy)
  • Sort the methods in alphabetical order to remove the pro-CSSD bias
  • Remove Majority Choice Approval altogether (I believe that MCA is original research)
  • Add Ranked Pairs, which has a Wikipedia article and seems to be on an equal level of notablity with CSSD
  • Add majority criterion and consistency criterion (which needs an article)
  • Add Independence of irrelevant alternatives, including local independence
  • Add independence of clone candidates. Possible values for the column are "Yes", "No (teaming)", "No (vote-splitting)", and "No (crowding)". Teaming trumps crowding.
  • Collapse "generalized" or "weak" criteria into the main criterion - they rarely make a difference

When I had to add cells to the table, I looked up the information on condorcet.org or Electowiki. I apologize if I made a mistake in any of the cells. Also, if someone can fill in the "unknown" information for Ranked Pairs, it would be quite helpful.

RSpeer 03:28, Apr 12, 2005 (UTC)

Thank you to the anon at 62.246.* - your changes do seem to be correct. I hastily reverted them (having my guard put up by seeing anonymous changes with no edit log), then realized that I had screwed up originally and the anon was right. RSpeer 14:08, Apr 12, 2005 (UTC)

Table

I don't want to edit myself but suggest the table and criteria are a little confusing in that the criteria are generally expressed in a negative formulation, but the table appears to be positive about that negative, if you see what I mean. May I humbly and tentatively suggest re-wording the criteria to a positive sense, if only as 'avoids the problem of ...'.

--Chuck (2005-05-12)

You don't need to be so humble - I was just blatantly inconsistent in writing those criteria. Is it better now? RSpeer 23:08, May 10, 2005 (UTC)

Yes, a lot clearer. -- Chuck (2005-05-13)

electowiki link

F451: If you go to this page http://wiki.electorama.com/wiki/Special:Allpages you will find various articles that exist on electowiki but do not exist on wikipedia. Certainly there is a lot of duplication, but the extra material makes it worth an external link. Hermitage

It's another secondary resource, but in this case, mostly an older, obsolete version of the wikipedia articles that have since been edited to much higher quality. Let's cultivate a high standard of scholarship, James. Anything that is there and valid should be copied over to wikipedia and edited here. There is a worthwhile project for whoever wishes to do it.--Fahrenheit451 00:58, 30 May 2005 (UTC)

  • I disagree. Electowiki serves a purpose, because ongoing research in election methods can go there, while it can't go on Wikipedia until it's more established. We should link to it. Your complaint seems to be that, on the topics where it overlaps with Wikipedia, it hasn't been obsessively edited to keep in sync with Wikipedia. That doesn't mean it's worthless: that means it has fewer editors, plus merging content from one GFDL resource to another can be confusing and difficult. RSpeer 21:33, May 30, 2005 (UTC)

You are falsely alleging that I require "obsessive editing to keep in sync with wikipedia" and I think electowiki is worthless. Please present some examples of innovative, ongoing research on electowiki. Otherwise, there is no reason to link to it.--Fahrenheit451 02:01, 31 May 2005 (UTC)

Why were you unable to check out the site yourself (as I requested) before you deleted the link? Hermitage

Because I am not a student on summer break like you are and have lots of time on my hands. I have a career and other responsibilities. You are perfectly capable of presenting a case of the inclusion of a dubious link.--Fahrenheit451 13:43, 31 May 2005 (UTC)

I've said before that if you don't have the time to read a source, you don't have the time to remove that source as an external link. Hermitage

I wonder if your comment is elitist. Sure comes across that way. I looked over some, but not all, of the articles, and found nothing of any value. Perhaps a voting methods research article should be created and electowiki would be an appropriate link. I will continue to remove the electowiki link for the reasons I have stated.--Fahrenheit451 19:32, 1 Jun 2005 (UTC)

It would have been very simple to go to the link I posted above and notice that there are several articles there that are not in the wikipedia.

Wrong. I have no idea what is different between the two. I did ask for examples.--Fahrenheit451 13:43, 31 May 2005 (UTC)

Okay, well, now you know. Hermitage

However, I have taken some more of my own time to copy the names of some of the articles below. Some of these may be in wikipedia, but I believe that the majority of them are not.

Which ones are and which are not?--Fahrenheit451 13:43, 31 May 2005 (UTC)

As far as I know, none of the articles below are in wikipedia, but I can't say that for sure. Hermitage

As for your comments about "high standards of scholarship": 1. It smells of elitism. Hermitage

Wrong again. That is exercising responsibility for the standards of wikipedia content.Evidently you don't know what the word elitism means.--Fahrenheit451 13:43, 31 May 2005 (UTC)

If you assume that only people whose articles are published in journals have something useful to say, that's elitism. I'm sure that your actual views aren't this extreme, but they seem to tend in this direction, i.e. your insistence that external links should be published and peer reviewed. Hermitage

Wrong again. The publication and peer-review process gives a particular professional community the opportunity to comment on a scholar's work. It is an imperfect, but valuable process.--Fahrenheit451 19:32, 1 Jun 2005 (UTC)

Who are you to decide what the standard of scholarship should be, and whether a given site meets those standards? Hermitage

Because I am a wikipedia editor.--Fahrenheit451 13:43, 31 May 2005 (UTC)

Right, well, we're all wikipedia editors here, so that's not really saying much. My point is that you don't have a right to decide these things unilaterally. If your standards of scholarship are not embraced by a broad consensus of wikipedia editors, then you don't have a right to impose them on any given page. Hermitage 21:55, 31 May 2005 (UTC)

Wrong again, James. I have a right to edit in accordance with wikipedia policy and I exercise those rights fully.--Fahrenheit451 19:32, 1 Jun 2005 (UTC)

That's it? That's your whole reply? You asked for examples of original research unique to electowiki, and I gave you 24 examples. Yes, electowiki duplicates out of date wikipedia articles, but it also has original work which makes it easily as worthwhile as some of the other pages linked to this page. My guess is that you're deleting the link for some sort of emotional reason, e.g. to avoid losing face. Anyway, your reply seems to contradict the spirit of wikipedia. ("Wikipedia works by building consensus.") I haven't read up on the dispute resolution procedures yet, but for now I'm going to hope that someone else besides me adds the electowiki link, so that F451 sees that I'm not the only one who thinks that it has merit. Hermitage 23:07, 1 Jun 2005 (UTC)

Descending Acquiescing Coalitions

Direct Democracy via Proxy Voting

Imagine Democratic Fair Choice

Later-no-harm criterion

Marginal Ranked Approval Voting

Neutrality of Spoiled Ballots

Pairwise-Elimination

SCRIRVE

Approval Sorted Margins

CDTT

Cardinal pairwise

DSC

Descending Acquiescing Coalitions

MMPO

Median Ratings

Pairwise Sorted Approval

Pairwise Sorted Dyadic Ballots

Ranked Approval Voting

River

Sprucing up

Uncovered set

Definite Majority Choice

Descending Solid Coalitions

Vote For and Against

Hermitage 12:15, 31 May 2005 (UTC)

Challenge to validity of Independence of Clones criteria

I challenge the validity of the Independence of Clones criteria: I have not found any mathematical definition of a clone, nor have I found a way to objectively determine what a clone is and isn't. I can say that clones are a special case of a correlated pair, but that still is not a precise definition. I would like to promote a discussion on this topic.--Fahrenheit451 18:29, 6 Jun 2005 (UTC)

  • Independence of clones is a commonly-used way of formalizing the effects strategic nomination. Clones are defined precisely on the strategic nomination page. Deleting a column from a table is a very poor way of "promoting discussion". What was your reason for doing this? RSpeer 04:51, Jun 7, 2005 (UTC)

There is nothing precise about that definition, furthermore, show me some actual clones.--Fahrenheit451 20:00, 7 Jun 2005 (UTC)


Independence of clones is mathematically well-defined. See T.N. Tideman (1987): Independence of Clones as a Criterion for Voting Rules. Social Choice and Welfare, 4: 185-206

I am aware of that reference and cited it in a past edit.--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

You are correct in saying that perfect clone sets are unlikely to occur in public elections. However, the criterion is still useful for analytical purposes. Also, methods failing clone independence are more likely to respond irregularly to the addition or subtraction of candidates who are close to being clones of other candidates.

There is no evidence that any clones have ever existed. When you say useful for analytical purposes, you are giving your opinion, which with I disagree, but you are also implying icc is nothing more than a theoretical construct. I say it is not well-defined because it does not identify observable phenomena, and it does not lead to the prediction of phenomena yet unobserved. I would go so far to assert that the independence of clones criteria is Junk Science.--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

Independence of clones is a criterion that many students of voting theory find to be significant.

Exactly which students are you refering to?--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

Thus, I see no reason to delete it. If you find the criterion unimportant, that's fine, but there is no need to impose that judgement on all wikipedia readers new to the topic.

I have no interest in deleting the article, just removing the references to it in relations to various voting methods. I can invent a criteria, call it the independence of the tooth fairy criterion, and edit the various voting method articles to state which ones comply or don't comply with itfc. Even if some editors "find it useful" it is absurd to use it if the tooth fairy can never be observed or proven to exist.--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

Remember, Wikipedia works on the basis of consensus. If there is wide agreement that ICC is at least worth considering as a voting systems criterion (which there is), then it should be defined on wikipedia, and it should be listed among the criteria passed and failed by the various tally methods. If all wikipedia editors made a point of deleting criteria that they didn't find useful, there would be few to zero criteria left in the encyclopedia. Please don't be so contentious in your editing.

Wikipedia also requires content to be verifiable, and these "clones" are not. It is irrelevant if 100 wiki editors proclaim the tooth fairy to exist when the existence of even one can never be verified or proven.--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

With regard to ICC, you seem to be largely editing before discussing, rather than the other way around. If you are having trouble understanding why people find merit in the ICC, I suggest that you discuss more before editing. In addition to discussing on wikipedia, I invite you to discuss these issues on the electorama mailing list. Go ahead and make an anti-ICC argument there, and see what kind of replies you get. That should be more productive and less antagonistic than the current pattern of "edit wars" here on wikipedia.Hermitage 04:54, 8 Jun 2005 (UTC)

I did browse the postings to that list. To me, most of it is noise rather than signal. The opinions on that list do not interest me, rather I am interested in verifiable facts. Tideman's "clones" are Not verifiable.--Fahrenheit451 15:27, 8 Jun 2005 (UTC)

Voting method criteria are not facts, they are mathematical definitions.

Clone is not a criteria, but is presumed to exist in Tideman's definition. If clone candidates do not exist, then discussing them is tantamount to discussing fiction.--Fahrenheit451 18:56, 8 Jun 2005 (UTC)

If you do not understand what a mathematical definition is, you do not have the necessary background to write about criteria.

You are in no position to evaluate my mathematical aptitude. Just to clue you in, I pulled straight A's in calculus in high school and only attended class to take the tests. I also passed the final exam for first year college chemistry with calculus prior to my high school senior year and took organic chemistry and lab while still a high school senior. Have a good day, Rob.--Fahrenheit451 18:56, 8 Jun 2005 (UTC)

"Clones" are a part of this definition, not something whose "existence can be verified or proven". Here are other similar Wikipedia articles, about idealizations that "do not exist" but are still useful definitions:

RSpeer 16:27, Jun 8, 2005 (UTC)

Yes, and they are totally different bodies of data than clones. Those are not people, for one, and for two, absolutes are unobtainable in the physical universe. The three data you cited are absolutes. Clones are a theoretical construct for some imagined candidates who statistically correlate on ballots. The error in your thinking is that you are equating mathematical absolutes with the much larger superset of theoretical constructs. "Clones" are Not well-defined as they do not identify any observable entity. Hence, are of no practical use.--Fahrenheit451 18:56, 8 Jun 2005 (UTC)

Well-defined is a mathematical concept, and clones certainly qualify as a well-defined object - there is no ambiguity in the definition as to whether something is a true clone or not. Whether or not they are likely to be commonly found in elections involving people is entirely secondary - just as no real-world temperature will ever be absolute zero, it is still a useful definition for analytical purposes. As was mentioned earlier, analysis on the effect of clones is frequently useful as analysis on the effect of near-clones as well. Clone analysis, for instance, can be used to study what the effect of putting the same candidate in two different places on the ballot would be. Clones are not a concept invented on Wikipedia either - they appear in the voting literature with some regularity. Scott Ritchie 19:46, 8 Jun 2005 (UTC)

Please cite where the ICC appears with regularity. Please substantiate your statement.--Fahrenheit451 18:40, 12 Jun 2005 (UTC)

Am I right in that what you are saying is that clones are an interesting mathematical concept to you and some other folks, even though there is no evidence for their existence? I just do not agree that voting criteria should have no real world application. It is not number theory where such endeavors are legitimate. I have no objection to anyone playing with theories, but please do not mix that data in with data that does have real application with verifiably extant objects.--Fahrenheit451 00:30, 10 Jun 2005 (UTC)

There's no evidence of the existence of ideal gases either, however they are a very useful concept for studying real-world chemistry. To claim they have "no real world application" because empirical research shows them to be only very good approximations rather than a perfect model is...strange. True clones may not commonly exist in large public elections (just as true ties don't), however near ties and close elections certainly do, and theoretical analysis of their effects is certainly done - hence, it belongs in Wikipedia. Scott Ritchie 01:30, 10 Jun 2005 (UTC)

Fine, so show me an article where a real candidate election was analyzed using ICC.--Fahrenheit451 18:43, 12 Jun 2005 (UTC)

I mentioned this elsewhere (unfortunately this discussion spreads over several pages), but I'll repeat it here: The category we're dealing with is called "voting theory". Thus, it is not limited to observation of real elections. Hence the definition of clones can be useful here even though clones may never exist in large public elections.

You ask which students of voting theory find the ICC to be interesting. There are far too many to list. Do a lit review. Do a google search. It's not our responsibility to spoon-feed this information to you. If you do the necessary research, you will find that ICC is a widely-discussed criteria, as far as voting criteria go, and therefore worthy of attention in wikipedia. If you don't have time to do the research, then you don't have time to delete ICC from wikipedia.Hermitage 23:15, 8 Jun 2005 (UTC)

I wonder if Fahrenheit's talk about the Independence of Clones Criterion could have anything to do with his advocacy of a certain election method ;)- Written by a anonymous coward with the i.p. address of 84.144.84.53

I wonder if Hermitage's talk about Independence of Clones Criterion has anything to do with his use of it in a derogatory article on the Borda count he has on his college webpage that he is not able to get published and peer-reviewed anywhere.--Fahrenheit451 18:36, 12 Jun 2005 (UTC)

NPOV notice

I can't tell when this page had anything POV on it recently, regardless of whether you think independence of clones is a good criterion. This page doesn't express any opinions about criteria, it just lists them. Can we remove the NPOV notice?

RSpeer 02:44, Jun 10, 2005 (UTC)

I think so. Perhaps the npov flag served to foster discussion about ICC, but I think discussion can continue without it. I don't think that the inclusion of a criterion like ICC is a sufficient reason to label a page as non-neutral. Hermitage 07:13, 11 Jun 2005 (UTC)

New mechanism for adding and removing methods/criteria

I'd like to propose a page where we discuss adding and removing methods or criteria from this page and its table. This is due to the large amount of editing that is involved in making such a change - it would be best to achieve consensus first. Please add /Included methods and criteria to your watchlist.

I've started that page with a (perhaps surprising) proposal to add Majority Choice Approval as a method.

RSpeer 05:15, Jun 13, 2005 (UTC)

Inconsistent use of voting method criteria

Today I edited several method articles to bring the criteria listed to consistency. One method satisfied the foo criteria, but in another article the foo criteria was not listed, but the boo criteria was. For example, I removed the Smith criterion from the Borda count article because most articles did not use it, and Hermitage reverts my edit. We need a consensus on which criteria is used to evaluate methods and stay consistent with that in editing.--Fahrenheit451 23:59, 14 Jun 2005 (UTC)

I agree that consistency in this area could be beneficial. I will follow Rspeer's lead and continue the discussion of which criteria to use on the /Included methods and criteria page. Hermitage 01:00, 16 Jun 2005 (UTC)

approval/range and ICC

I altered the table to show that approval and range fail ICC. If you disagree, please make your argument here. Here is a failure example for approval:

Preferences with approval cutoffs:
60: A>R>>S
40: R>S>>A
Approval scores:
A: 60
R: 100
S: 40
R and S are clones, assuming that clones are defined by voter preferences rather than voter rankings. If you apply ranked ballot criteria to approval voting in the latter ("votes-only") way, approval passes virtually every criteria (Condorcet, Smith, mutual majority, you name it). Hence, saying that R and S are not clones would contradict several other statements on wikipedia about approval's status with respect to these criteria.

R and S are NOT clones, as 60 voters drew the distinction between R and S by approving of R but disapproving of S. If they were clones, then an approval distinction would not be drawn between them. This statement has no bearing on any other criteria. Hermitage's statement is badly flawed and false.--Fahrenheit451 00:14, 10 September 2005 (UTC)

R is the initial winner. However, if you remove candidate R, the winner is A.

It is even easier to show that range voting fails ICC, because you don't have to delve into the irritating preference criteria versus votes-only criteria discussion

Ratings
60: A100 R99 S0
40: R100 S99 A0
Rating sums:
A: 6000
R: 9940
S: 3960
R and S are clones. R is the initial winner. However, if you remove candidate R, the winner is A.

R and S are not clones. In the first profile, S is disapproved. A distinction is drawn between R and S there. --Fahrenheit451 22:12, 19 September 2005 (UTC)

So, I'm fairly well convinced that approval and range fail ICC. However, I'm less sure about whether the effect should be called "crowding" or "teaming". If you make the simplifying assumption that voter's cardinal/approval ratings of candidates are independent of the other candidates in the race (this assumption is IMO implied by the statement that approval and range pass IIA), then I think that it is purely a teaming effect. --Hermitage 09:57, 25 Jun 2005 (UTC)

You rule out "vote-splitting" for some reason. KVenzke
I refer you to the strategic nomination page, where these terms are defined. As I understand them, methods where nominating more clones is always a disadvantage to the clone set suffer from vote splitting. Likewise, if it is always an advantage, the method suffers from teaming. If it is sometimes an advantage, and sometimes a disadvantage, the method suffers from crowding. --Hermitage 28 June 2005 07:31 (UTC)
Then crowding it is. I can't seriously imagine a political party deliberately entering two candidates into an Approval race. KVenzke June 28, 2005 21:07 (UTC)
I can. Perhaps one will succeed where the other fails, i.e. at getting sufficient votes from outside the party. --Hermitage 21:04, 2 August 2005 (UTC)

Then they are NOT clones. You shot down your own argument.--Fahrenheit451 22:12, 19 September 2005 (UTC)

If you assume that voters approve or disapprove candidates independently, the addition of a clone can only help its clone set. This is a simplifying assumption. In real life, it is possible that the addition of a clone could cause voters to disapprove other candidates in the clone set whom they would otherwise approve. If so, then the effect is indeed crowding. However, as I argued above, if we do not take this simplifying assumption, I think that we should also say that approval fails IIA for practical purposes. Do you understand this argument? --Hermitage 29 June 2005 01:58 (UTC)

Rather than that or "crowding" or "teaming" I prefer Woodall's Clone-Winner and Clone-Loser. KVenzke

I'm not familiar with those. --Hermitage 28 June 2005 07:31 (UTC)
Approval would fail both, if you use an "arbitrarily placed cutoff" interpretation of Approval.
Woodall lists Approval as satisfying both. I'm increasingly inclined to use his interpretation of Approval, according to which "approved" candidates are ranked and "disapproved" ones are unranked. The reason why this appeals to me is that it seems to me that the only reason Approval could be said to fail clone independence is due to the ballot format. Change the ballot format to a ranked ballot [such that ranked candidates are "approved"], and the argument seems to disappear. But the method is exactly the same.
Also, it seems to me that it may be unfair to Approval to note that the members of a clone set might not receive the same votes. In practice, it is not likely that Schulze or IRV would see clones receiving the same votes, either. We don't count that against them because the clones are theoretical. That is no less the case under Approval. KVenzke 17:00, Jun 25, 2005 (UTC)
This last paragraph doesn't make sense to me. --Hermitage 29 June 2005 01:58 (UTC)
Woodall criteria aside, are you saying that approval voting passes the independence of clones criterion (the one defined by Nic Tideman)? If approval passes ICC, why doesn't approval pass Condorcet, Condorcet loser, majority, Smith, etc.? Depending on one's interpretation, I think that approval should either pass or fail all of these. IMO, it is absurd to say that it passes them all. I feel strongly that the table should not say that approval passes ICC. --Hermitage 28 June 2005 07:31 (UTC)
Why don't you read what I wrote above about the reason why this [interpretation] appeals to me? It's quite clear that this interpretation doesn't lead to Approval satisfying any of the other criteria you've listed.
The essential question is whether a rank method, in which the candidate receiving the most explicit rankings (the fewest truncations), is Approval voting. KVenzke June 28, 2005 21:07 (UTC)
I have read your interpretation of approval, but I don't think that it should be adopted on wikipedia. It is an unusual interpretation, which I think Tideman never considered when writing the ICC. In evaluating approval voting, Tideman offers two interpretations. (1) Ranked ballot, specifying a number of approved options. (2) Ranked ballot with only two slots. Given interpretation (2), approval passes all criteria. Given interpretation (1), it fails various criteria including Condorcet, Smith, and ICC. As I mentioned, I think that it is silly to use interpretation (2).
To be fair, Tideman does express some ambivalence about the ICC failure. He writes "The possibility that approval voting could generate the same votes as plurality also means that approval voting lacks independence of clones, but this failure to satisfy the formal definition overstates the vulnerability of approval voting to cloning. The definitions of clones and independence of clones were intended to capture the consequences of the presence of virtually identical options. If a set of options were truly indistinguishable, there would be no reason for any voter to cast an approval vote for one option in the set and not others, and they would all receive the same approval scores. In this sense approval voting should be regarded as independent of clones. However, if some voters make distinctions among very similar options, then approval voting lacks independence of clones just as plurality does." (This is from a manuscript that will soon be published, entitled "Collective Decisions and Voting".)
So, I would definitely not give approval a pass for ICC on the table. At the very best, it gets a question mark or something like that, but assuming that we don't want to pass out question marks, it should get a fail. --Hermitage 29 June 2005 01:58 (UTC)
So we're restricted to Tideman's interpretations of Approval voting?? I didn't realize that was his patent. By the way, you don't need to call it "your interpretation"; it is Woodall's interpretation. KVenzke June 29, 2005 15:32 (UTC)
Tideman wrote the ICC. We're not restricted to his opinion about whether approval passes ICC, but it serves as a logical default when we need to put one and only one value in the table.
Well, we agree that Tideman's interpretation (2) is useless. But it's not at all clear to me that interpretation (1) leads to Approval failing ICC. It just seems to beg the original question from scratch. KVenzke June 30, 2005 13:06 (UTC)
Anyway, I not sure that Woodall's interpretation is very applicable here. By his interpretation, do all members of a set need to be either approved by all voters or disapproved by all voters in order to qualify as a clone set? --Hermitage 29 June 2005 22:30 (UTC)
That's my understanding. I can see that this might cause Woodall's interpretation to make no difference.
A set of candidates is called a set of clones if every voter who votes for any of them votes for all of them, and in consecutive positions on their ballot. To clone a candidate x means to replace x by a set of clones on every ballot containing x. This concept was introduced by Tideman ["Independence of clones as a criterion for voting rules", Soc. Choice Welfare 4 (1987), pp. 195–206] and refined by Zavist and Tideman ["Complete independence of clones in the ranked pairs rule", Soc. Choice Welfare 6 (1989), pp. 167–173]; however, in [these], the definition of clones requires that every voter should express a preference between every clone and every non-clone, which we do not require. Douglas R. Woodall, "Properties of single-winner preferential election rules I: impossibility theorems", draft (2003). KVenzke June 30, 2005 13:06 (UTC)
I'm entering this discussion a bit late, but: teaming? If Approval has anything, it's vote-splitting, because voters would get tired of approving a long list of identical candidates.
Hermitage's analysis depends on making up unvoted preferences, and applying the clone criterion to those preferences instead of to the actual votes. I realize this is a long-standing debate, but to me this seems as fishy as Saari constructing a model of Condorcet as being Borda plus "Condorcet noise". You can make any method look bad by portraying it as a defective version of another, and this goes for portraying Approval as a vote where everyone first comes up with preferences and then wedges them into two slots.
Anyway. I realize that the clone criterion is badly-defined on range methods, and you could make them pass or fail depending on your interpretation, so we can't authoritatively say that Approval (and range voting) pass. But for the same reason, we can't authoritatively say that they fail. I think these boxes should be gray and say "not applicable".
RSpeer 20:55, July 12, 2005 (UTC)
I think rather than "not applicable" we should suggest that it depends on one's interpretation, or at least, that it is disputed. I change my mind; I think it is perfectly reasonable to say that ICC does not apply to methods which lack a ranked ballot (only) representation. So FPP is a fail; Approval could go either way (since more than one representation is possible); MCA (etc.) and range voting are N/A. KVenzke 21:07, July 12, 2005 (UTC)
My opinion remains that it should be listed as "fail", although I would certainly prefer a "non-applicable" or "ambiguous" designation to a "pass" designation.
If it is a fail, I still think that "teaming" is the best designation. To repeat my argument from above: If you assume that voters approve or disapprove candidates solely on their own merit, then approval passes IIA. If you do not make this assumption, than approval does not necessarily pass IIA. If you make this assumption, then adding a new candidate will not cause voters to disapprove another candidate. Thus, the addition of a new candidate representing an existing point of view can only help it; it cannot hurt it.
Of course, this assumption is not valid in the real world (and thus approval is probably not independent of irrelevant alternatives in the real world), but nevertheless I suspect that ideological factions will have a fairly strong incentive to nominate more rather than fewer candidates under approval voting, in the hope that one may succeed (to gain enough votes from outside the faction) where the others fail.
Back to the question of whether approval should be said to fail ICC. First of all, can we agree that range voting with a large number of slots fails ICC? If so, it seems silly for range to fail, but approval to pass simply because it doesn't allow voters to give expression to multiple levels of preference, hence casting the notion of multiple preference levels into the realm of the metaphysical. --Hermitage 08:25, 13 July 2005 (UTC)
By the way, I'm not arguing this because I dislike approval. Actually I'm lately been gaining a new appreciation of it. It's just that I don't think that it makes sense to say that it's independent of clones. Please consider the argument above before reverting my edit. Should we really give approval an ambiguous rating while giving range a straight fail? Can you make a credible argument that range should not get a straight fail? --Hermitage 09:14, 14 July 2005 (UTC)

I don't find your argument for a "teaming" designation to be persuasive. Running multiple candidates possibly will gain some votes from outside the faction. But it will almost surely split the votes from within the faction. I guess it's a question of whether we want to describe Approval's properties in practical terms (in which case Approval's problem is easily vote-splitting; but by the same reasoning, cloneproof methods suffer from vote-splitting as well), or in theoretical terms (in which case Approval satisfies IIA and possibly is vulnerable to teaming). KVenzke

Yes, I agree that it depends on whether we choose a practical or theoretical approach. Given a practical approach, it may be fair to say that approval suffers from crowding (more clones sometimes helps, sometimes hurts) and it should not be said that approval passes IIA. Given a theoretical approach, approval passes IIA and suffers from teaming. --Hermitage 07:49, 28 July 2005 (UTC)

Regarding range voting: Approval can be said to satisfy ICC if we treat it as a ranked method with ranked candidates approved, and use Woodall's notion that a voter who votes for (i.e. ranks) any member of a clone set votes for all of them. This argument can't be adapted to range voting. KVenzke 15:30, July 14, 2005 (UTC)

I thought that I already pointed out that this interpretation of ICC to approval is very weird. Candidates can only be clones in this interpretation if they are either approved by all voters or approved by zero voters, correct? --Hermitage 07:49, 28 July 2005 (UTC)
Substitute "voted for" for "approved" and you have part of Woodall's definition of a clone set. If {a,b,c} are clones, then all voters either vote for all of them or none of them. I believe your argument that Woodall's interpretation of Approval is "weird" is based simply on the fact that it never occurred to Tideman. KVenzke 01:03, August 1, 2005 (UTC)
No, not at all. For each voter, that voter either approves everyone in the clone set or disapproves everyone in the clone set. RSpeer 14:29, July 28, 2005 (UTC)
That would be a standard votes-only interpretation of approval. I was under the impression that Woodall's interpretation was somehow different. --Hermitage 21:22, 2 August 2005 (UTC)
Can't it? I'd say that, in a votes-only interpretation of ICC as applied to range voting, a voter gives the same rating to everyone in a set of clones. So I'd say the criterion is ambiguous for range voting too. I am less adamant that Range/ICC should be "Ambiguous" than I am about Approval/ICC, and I would not complain if someone changed it with a justification, but right now I can't see how range and approval can have different statuses. RSpeer 04:44, July 28, 2005 (UTC)
The interpretation I suggest for Approval (which allows it to be clone-independent) won't work for range voting, since a range ballot collects information that can't sensibly be collected by a rank ballot. Therefore it seems to me that range voting can't satisfy an ICC defined for rank methods, whereas Approval can, since it is possible to implement Approval as a rank method which satisfies rank ICC. KVenzke 01:03, August 1, 2005 (UTC)
People seem to confuse what they think ICC should mean with the way it has actually been defined. Most people think that clones should be candidates such that all voters are indifferent between them. This is clearly not the definition published by Tideman. If it was, methods such as minimax would pass ICC if they allowed for equal rankings. The definition of clones, counterintuitive as it may be, is that clones are candidates such that they always appear consecutively in voters rankings. The logical application of this for non-ranked methods is that they are candidates who always appear consecutively in voters' preference orderings. In my range voting example above, it is clear that R and S always appear consecutively in voters' preference orderings. Since the removal of candidate R costs the clone set its victory, range voting fails ICC.
KVenzke has already rejected the idea of giving Nic Tideman the last word on this subject (despite the fact that he made up the ICC), but anyway I'll mention that he has a table of criteria compliance in his upcoming book, and that approval voting is currently listed as failing ICC. --Hermitage 07:49, 28 July 2005 (UTC)
I'm not saying anything about what ICC should mean; it seems you are, given your range voting example. The issue is how Approval is interpreted. I don't agree that the "logical application of [ICC definition] for non-ranked methods is that they are candidates who always appear consecutively in voters' preference orderings." If a method is not a rank method then you can't expect to apply a criterion written for rank methods. So I say that range voting, which can never be a rank method, fails. (Notice if you run a Schulze election on rated ballots, we don't have this problem.) But Approval can be defined as a rank method, in more than one way, though not with the same result regarding ICC compliance. It seems very premature to me to say that Approval fails ICC due to the appearance of the ballot. KVenzke 01:03, August 1, 2005 (UTC)
Criteria can be defined in terms of voters' hidden preferences, or they can be votes-only. We know how the hidden-preferences ICC is defined: you suppose that the voters rank the candidates, and then choose an arbitrary cutoff between any two consecutive candidates. (I'd argue that no voter would put their cutoff between two clone candidates, but whatever.) What I'm talking about is the votes-only version of ICC. RSpeer 14:36, July 28, 2005 (UTC)

We have had a discussion previously about tideman's icc. I suggest that if it were better defined, these discussions would have been obviated. The key to a better definition, in my view, is to call it independence of correlated candidates. I have to agree with Kevin's statements on the subject and regard them as superior to tideman's.--209.4.43.13 21:17, 30 July 2005 (UTC)

My opinion is that if approval passes ICC, then it also passes all the other criteria on the table. If we have no consensus about how to apply ranked ballot criteria to approval and range, we should probably just take it off the table. However, I do somewhat like the "ambiguous" designation as long as it links to this discussion page. --Hermitage 07:40, 2 August 2005 (UTC)
In my opinion, the interpretation of Approval which results in it satisfying Condorcet and Majority is not supportable, since it would work just as well for Plurality. I don't object to the "ambiguous" designation. But it is an interesting, important problem, that established criteria are written for rank methods, but not all methods are rank methods. KVenzke 15:38, August 2, 2005 (UTC)

Ley de Lemas

I'm trying to describe a voting system that is used in some provincial and municipal districts of Argentina, but I can't really make sense of it in terms of the types mentioned in this article (probably because the system itself doesn't make much sense). In the province of Santa Fe it was adopted under a law called "Ley de Lemas". I don't know what algorithm it employs to assign winners; my issue is with how the votes are counted.

  • Each party (called a lema) may present several candidates or lists of candidates (sublemas) for each post to the main election (avoiding the need for primary elections).
  • Citizens vote for a sublema within a lema (the ballot says "Party X, Candidate Y").
  • The winning sublema is the one which got most votes within the winning lema. The winning lema is the one which got most votes (the cumulative of all its sublemas).

The last point means that a candidate can win an election having obtained far less votes than his rival, and many of these votes may be in fact for other candidates in the same lema.

The Peronist party employed this system to present a "supermarket"-type display of candidates in several elections. In some cases there were 20 or more candidates for the same post, most of them insignificant, but all of them contributing to the party's vote pool. In 1995 the party went to the Santa Fe elections with several candidates for governor, against one powerful rival of the Socialist Party. The two main Peronist candidates got 21 and 18% of the votes respectively (totalling 39% for the party as a whole); the Socialist got 33%. The Peronist with 21% of the votes won the election. Does this system even have a name, or should it be called a scam pure and simple? --Pablo D. Flores 6 July 2005 12:40 (UTC)

The mechanism itself isn't that strange, and it can be described in terms of the methods we have here. It's open-list party-list proportional representation with one winner. The strange part is that party-list PR is designed as a multiple-winner electoral method, and Argentina is using it to elect one winner, which takes away the "proportional representation" part of the method.
So, the system is odd, but many other electorates have even odder systems and they tend not to be considered scams. In this system, if voters are assumed to have preferences for parties and not candidates (a strong assumption, but all party-list PR methods make it), then the candidate who is elected is at least a member of the party preferred by the plurality. So it's a lot like plurality with a primary, in fact, except that voters don't know the result of the "primary" (the sublema winner) when they cast their vote.
The issue you raise, I believe, is that the 18% who voted for the second Peronist candidate may not necessarily like the other candidate who got 21%, and if they had known their vote would go to him they would have voted differently. Right?
Hope this helps. RSpeer July 6, 2005 14:46 (UTC)
Thanks a lot. Yes, that's exactly the issue. In Argentina most party-based ideological platforms vanished in the last decade, as parties decided to give up convictions and focused on simply clutching to power at all costs, so most people vote for candidates. The Peronist party in particular has a problem with intra-party diversity — it accepts all of it, so this Ley de Lemas is perfect for them (it allows them to get votes from the whole ideological spectrum). Here in Santa Fe it was abolished this year (after much pressure); in other places it still holds. I'll see if I can integrate all that into an article now. Thanks again. --Pablo D. Flores 6 July 2005 15:03 (UTC)
RSpeer's comment is helpful. This seems like an "open list" system applied to single winner elections. (In closed list, you vote only for a party. In open list, you vote for a party and also for a candidate within the party.) Curious! --Hermitage 7 July 2005 06:21 (UTC)


Table is Biased POV towards Condorcet methods

It is very interesting that Condorcet methods fail IIA and a special, similar named definition has been contrived to pretend to make them satisfy IIA. This has tainted the voting methods table and was removed by myself, then later reverted by Hermitage. This is makes the table clearly biased towards condorcet methods.--Fahrenheit451 9 July 2005 15:14 (UTC)

If this table was "biased towards Condorcet methods", then it would certainly not contain criteria like participation and consistency which are known to be incompatible with the Condorcet criterion. Markus Schulze 9 July 2005
Your premise is not correct Markus: One column title reads "independence of irrelevant alternatives". That is a specific criterion that ranked systems fail. Then, in the blocks for the two listed condorcet methods, ranked pairs and your method, a convenient exception is made for a different definition. That is POV and improper.
So, the table IS biased towards condorcet methods as long as that exception is made. Condorcet methods fail IIA and it should be factually noted in the table.--Fahrenheit451 9 July 2005 16:00 (UTC)
I am hesitant to remove local IIA from the table not because of bias, but because I think that it is a fairly interesting concept that will not get much exposure otherwise. However, I can understand F451's point, and I don't feel strongly about the issue. --Hermitage

I agree that local IIA is interesting, as well, and would get exposure when someone looks into the more sophisticated condorcet methods, like ranked pairs and Schulze's method.--Fahrenheit451 17:45, 12 July 2005 (UTC)

I would have agreed with User:Fahrenheit451 some time ago. This page (as of now) is not as bad as it used to be, but it used to be aimed at criteria which required or promoted Condorcet methods. Many of the criteria articles are still biased to Condorcet methods. To pick one a random: Mutual majority criterion is clearly designed to justify Condorcet methods. --Henrygb 17:22, 10 July 2005 (UTC)

Eh? The Mutual majority criterion is also satisfied by Instant-runoff voting. Douglas Woodall has written suggestions that it's the among the most important criteria (since it is the single-winner case of the "Droop proportionality criterion"), and Woodall doesn't give the impression of advocating Condorcet methods. KVenzke 00:02, July 12, 2005 (UTC)
Furthermore, some Condorcet methods fail mutual majority! For example, minimax, Black, Dodgson, etc. Condorcet plus clone-proofness implies Smith compliance. Smith compliance implies mutual majority compliance. --Hermitage 05:41, 12 July 2005 (UTC)

The last vestige of the Condorcet method bias in the table is the local IIA designation in the blocks for the two list condorcet methods, ranked pairs and Markus Schulze's method. All ranked methods fail IIA, so it is flagrant bias to stick a reference to another definition that satisfies the two mentioned condorcet methods in the table. Perhaps Markus is being inordinately defensive of his method, but I think the merits of it stand on their own. Two other editors are also quite condorcet partial as evident from the methods they advocate in and off Wikipedia. I have even had people I know email me about the pro-condorcet bias on Wikipedia. It was a long process to achieve a consensus on the Borda count article, which was quite horribly anti-Borda six months ago. It is interesting that pro-Borda folks do not embark on campaigns against condorcet methods, but pro-condorcet people seem more than eager to suppress Borda count methods. --Fahrenheit451 17:40, 10 July 2005 (UTC)

I'm not sure what arguments Borda advocates would use against Condorcet methods. The anti-Condorcet arguments I can think of all apply doubly to Borda. KVenzke 00:02, July 12, 2005 (UTC)
I guess you have proven my point. Borda advocates usually don't attack condorcet methods. The flaws are not double as you state, but rather different. Anyway, the discussion has gone off-topic.--Fahrenheit451 01:04, 12 July 2005 (UTC)
Well, I had in mind the criticism that in Condorcet methods, strategic and counter-strategic voting can result in the election of a candidate whom nobody likes. Also, there's the criticism that the winner of a Condorcet method need not be any voter's first choice. These criticisms both apply also to Borda. KVenzke 20:57, July 12, 2005 (UTC)

You are correct on that point. Reflecting consensus is an important factor in any useful voting method in my view. I have actually devised a Borda variant which, in part, mitigates the phenomena you have pointed out here. If you are interested, I could send it to you.--Fahrenheit451 17:54, 13 July 2005 (UTC)

If you have invented a new method, I would suggest that you post it to discussion lists such as the election methods list at electorama. This gives others an opportunity to give feedback, plus it keeps a dated record of the proposal. --Hermitage 21:09, 13 July 2005 (UTC)
I am aware of that list. Have looked at a number of postings and find it much more noise than signal. I have sent the article for peer-review to some mathematicians. No interest in electorama at all.--Fahrenheit451 22:41, 13 July 2005 (UTC)
I was only trying to be helpful. Of course you are free to opt out of the EM list, but there is no need to belittle it. I personally have found it to provide an excellent learning experience. I am not interested in all of the posts, but it is often a good place to have questions answered and to clarify one's understanding of contemporary voting theory. Sometimes there is a fair amount of noise, but that is easy enough to ignore. Anyway, I would be happy to take a look at your Borda variant. --Hermitage 07:34, 14 July 2005 (UTC)

I would remind everyone of this article, which discusses "fairness" in journalism. It's not our job to make things look balanced; if IRV doesn't look so good in this table, there is not a need to correct the table to make it look more even, unless we're talking about adding more columns. Don't remove tests to make things look more even, add tests that are important, and if a voting method still fails a lot, blame the method, not the table. - McCart42 (talk) 17:27, 14 July 2005 (UTC)

I agree. Also, after further reflection, I think that local IIA is a fairly useful concept. Let's take the premise that IIA failure should be the formal definition of the "spoiler effect", and for ranked methods, the cause of the incentive for compromising strategy. If all viable ranked methods have IIA failure, then IIA is not good criterion for differentiating between ranked methods. The question then is under what circumstances a non-winning candidate can change the outcome. In Condorcet methods, this can only happen if there is a majority rule cycle. Thus, there is no compromising incentive if there is no majority rule cycle. In IRV, there will be an IIA failure if there is a majority rule cycle, but there may also be an IIA failure in the absence of a majority rule cycle. Thus it can be said with some mathematical rigor that Condorcet methods reduce the spoiler effect to a greater degree than IRV.
As to the Borda count, it may certainly have IIA failures even when there is no majority rule cycle. However, I'm not entirely certain that Borda count necessarily has an IIA failure in the case of a majority rule cycle. I'm 99% confident that a decent computer simulation would reveal more IIA failures in Borda than in Condorcet methods, but I don't know if this can be proven in a simpler and more definitive way, as with IRV. If I ever gain the ability to do good computer simulations, I will do simulations to find degree of IIA failure in different rank methods.
Anyway, back to local IIA. Basically, since regular IIA doesn't differentiate between viable rank methods, it makes sense to have an intermediate criterion which can do so. Local IIA seems like a somewhat sensible choice. Of course, it may make sense to have more than one intermediate criterion... --Hermitage 22:04, 14 July 2005 (UTC)

All ranked methods fail IIA, period. The table is not there for plastering different definitions in the blocks. I am continuing to remove this "local iia" flag from the blocks whenever it appears. I am putting that article on my watchlist. --Fahrenheit451 23:39, 15 July 2005 (UTC)

I understand your argument, but since there are different opinions on the matter, I think that we should put it to a vote. --Hermitage 00:47, 16 July 2005 (UTC)

We all have a right to our opinions, however you never compromise with the truth. You also do not vote on the truth. It will go this way: I refuse to participate in your "vote". The pro-condorcet faction will vote to stick "local iia" back in the two blocks, and I will continue to remove it from here on out. Tyranny of the majority is totally unacceptable to me.--Fahrenheit451 01:27, 16 July 2005 (UTC)

Please read the Wikipedia policy on resolving disputes. An edit war is not going to be a productive way to resolve this. Personally, I think the answer is to break out IIA and Local IIA into two separate columns. -- RobLa 06:22, July 16, 2005 (UTC)

Actually, a better solution is to have a seperate table for condorcet methods. The condorcet mavens can put whatever criteria they want in the columns without making it biased against other methods. There has been interest in Smith criterion and local iia which only certain condorcet methods satisfy. This way, the condorcet (winner) criterion can be removed from the non-condorcet method table, and the condorcet method table can omit participation and consistency criteria, which all condorcet methods fail. I think all parties would find this to be an acceptable compromise or favorable.--Fahrenheit451 18:59, 16 July 2005 (UTC)

How would you then compare IRV vs. Condorcet methods, or approval voting vs. Condorcet methods? If they are in two separate tables, doesn't that make comparison more difficult? I might like to see what it looks like before I judge though. F451, would you like to demonstrate in a userpage sandbox? - McCart42 (talk) 22:35, 16 July 2005 (UTC)

No it does not make comparison more difficult. You simply juxtapose one table over the other. Two tables will solve the criteria dispute issue. I created them mentally and comparison is quite easy, plus you can put condorcet specific criteria in the condorcet method table.--Fahrenheit451 22:48, 16 July 2005 (UTC)

Sorry, splitting into two tables is entirely unacceptable. It's a comparison table, and Condorcet methods need to be in the mix, not relegated off to a kiddie table. I'm thinking now that footnotes might be the way to go on this. So, it can say "No (passes local IIA)" or something to that effect. I think it's fair that the article should be clear that Condorcet methods fail IIA, but it's also important to highlight the counterargument. This would ideally be done for all methods and all criteria in the table. -- RobLa 23:00, July 16, 2005 (UTC)

I would go along with footnotes, but not comments. For example, NO (see 1.), but not, NO (passes local IIA), which is tantamount to the situation we had before. Local IIA is a different definition than IIA. I think footnotes would be NPOV and work.--Fahrenheit451 23:34, 16 July 2005 (UTC)

Whether to use numbers or words is a question of linking style, not NPOV. Numbering is harder to maintain, and provides less information. See this article on writing meaningful link text for more arguments on this front. Link numbering would make it harder to link to an external location (e.g. putting the caveats in a section of the offending method). I think its reasonable to insist that the word "No" and the pink color remain, but I'm hoping you can suggest something easier to maintain and more descriptive than a numbered list. -- RobLa 04:47, July 17, 2005 (UTC)

Point well taken. Then how about "NO (see local iia)"? This would make sense and does away with numbering difficulties.--Fahrenheit451 17:12, 17 July 2005 (UTC)

doubling of page content: be on the lookout for possible deletion

It seems that KVenzke's 20:57, 12 July 2005 edit unintentionally made it so the content of this page repeats twice. I will attempt to restore normalcy by deleting the top half. WARNING: This may delete new content. Please help me restore normalcy by attempting to find and restore new content that may be deleted in this edit. Thank you. --Hermitage 09:02, 14 July 2005 (UTC)

Sorry about the inconvenience. I was trying to make an addition, and it seemed that afterwards RSpeer's addition was lost. KVenzke 15:04, July 14, 2005 (UTC)


STV and IRV point of information

I noticed that the methods table states that irv satifies condorcet loser. This is only true if voters are not allowed to truncate ballots, as is the law in Australia. If voters are allowed to truncate ballots, and that is actually the standard form of STV and IRV, then the condorcet loser can be elected as other candidates may never receive the cl's transferred votes. This should be noted in the table. --Fahrenheit451 18:23, 5 September 2005 (UTC)

This is a half-baked idea. If you used this as the standard throughout the table, then most ranked methods would fail most criteria. For this purpose, the standard practice is to evaluate ranked methods with respect to actual rankings, not with respect to unstated preferences. --Hermitage 20:39, 5 September 2005 (UTC)

I would say your reply is half-baked: Voting methods are evaluated on basis of common rather than exceptional employment. In this case, a voter truncated ballot is standardly allowed in STV and IRV, the method used in Australia is the exception. If the preferences are unstated, they do not enter the tally. In the case of IRV, the standard method violates condorcet loser, whereas in Australia, the method satisfies Condorcet loser. --Fahrenheit451 22:54, 5 September 2005 (UTC)

Please clarify - are you saying that the same ballots could elect a winner in IRV who would be the Condorcet loser? And not, say, someone who would have been the Condorcet loser if ballots weren't truncated, which is what Hermitage's reply indicates? If it were the second, nothing would satisfy Condorcet loser, not even Condorcet, so it wouldn't be particularly meaningful. RSpeer 04:37, 25 September 2005 (UTC)