![]() | This page 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. |
Archives: ~ Jan 24, 2008 - Feb 13, 2008
An anonymous editor took out the entire San Francisco text, with a cryptic edit summary, which I'm sure will make sense after it is explained. There is a lot of data in that text, the editor did not explain what, specifically was wrong. But there is only one percentage reported as such, so if that alone was wrong, it should have been corrected, not removed, or, at least, not the entire report. I've reverted, and invite this editor or any other to find and explain errors. These facts were reported here before being put in the article and are fully supported -- except for any errors, of course -- by the source. So, please, if there are errors, *fix* them! It is not my personal responsibility, that is a misunderstanding of what Wikipedia is. It is not a debate forum. -- Abd ( talk) 15:33, 24 January 2008 (UTC)
header 1 | header 2 |
---|---|
data a1 | a2 data |
data b1 | xx data b2 |
Now that the U.S. impelmentation mess has been moved to a new article, I am not sure if this is the best page to explain some errors in Abd's analysis, but the new article doesn't seem to have a discussion page yet. So... One place where Abd has simply gotten it wrong with his "original research" is his reporting on San Francisco. He has misunderstood the way San Francisco reports the results. What the city does is treat people who skip over the race entirely as "undervotes". That means someone could win 55% of votes in the first round, but Abd would still call them someone elected with a minority of the votes. It is the standard, normal, and correct that percentages are always reported as a percentage of votes cast IN THAT RACE, rather than as a percentage of votes cast in other races (such as for president). The drop-off for lower offices is typical. The fact that most jurisdicitions use a combination ballot paper with multiple offices does not alter that correct way of reporting percentages.
Take the 2005 citywide race for city-assessor. Phil Ting won 47% of the first choice vote. He then won substantial number of second choices from another Asian candidate and won with an absolute majority of the first-round vote. But Abd has him as not winning a majority of the vote, because he counts in people who didn't vote in the race at all. See:
http://www.sfgov.org/site/elections_index.asp?id=61607
Similiarly, in 2006, Chris Daly won in the 5th round of counting with a majority of continuing votes. They stopped the count at that point with him something like 4 votes away from a majority of the valid first-count vote, so one could argue he technically didn't win a majority of all votes cast in the race, but they stopped the count with his three strongest opponents still in the race, because he had won a majority of continuing ballots (discounting exhausted) -- meaning he obviously would have won a majority of the first-round vote if they had reduced the field to two.
It is true that the hotly contested race in District 4 was won by a majority in the last round, but that majority was less than 50% of the valid first choice vote. That also has happened in a couple other Supervisor races, but people have accepted the results just as much as they would have if someone ran a runoff with the voter turnout a little lower in the runoff. It is never proper to describe the percentage in the runoff by using the denominator from the first round of voting.
Abd really messes up the race he profiles in the write-up - -the one with 22 candidates. He suggests the winner had only 37% of votes, but he skips over the MAJOR fact that the count ended with three candidates still in the race -- the winner had a majority of the final round vote with two other candidates still in the field so they didn't need to eliminate the third-place finisher, because no other candidate had any mathematical chance to win. So Abd is really comparing apples and oranges. True, Mirkarimi would not have won 50% of the first-choice vote if they had gone to the final two, but he almost certainly would have been a good bit over 40% -- and his point just is highly misleading and does not belong in a NPOV article. Tbouricius ( talk) 21:20, 26 January 2008 (UTC)
Abd. You might want to redefine the academic consensus that runoffs and instant runoff voting are majority voting systems, but you know you are doing so outside that academic consensus. Runoff elections also can have significant changes in turnout -- when one of two rounds can be decisive, significant differences in turnout obviously can raise questions about whether it was a "majority result", but the academic consensus is that it is. Choosing to make your case within the Wikipedia article that it's not is a highly political decisions, reflecting your own point of view.
In San Francisco, ithin the city, there was absolutely no controversy about the result of the Mirkarimi win in 2004. People were very used to candidates winning majorities in runoffs with a drop in turnout, and that's how this was seen --- with the added fact that the turnout in a real December runoff would have been much, much larger. Mirkarimi had a maority of continuing votes, people knew those were the rules, and he was generally seen as winning a majority in the instant runoff because he was the majority choice of the top candidates. You obviously can dispute that based on your own numbers, but it's one definition of majority versus another.
Furthermore, your selective highlighting of the 37% number in a short summary of data is highly misleading when there were three candidates left. Did you not know that when you did this or chose to overlook it?
As a final point, every winner in the thousands of election in Australia's house of representatives has won with an absolute majority of the valid first-count ballot. That's a product of their rules, which invalidate ballots that don't rank everyone. So if one wants absolute majorities of first-round ballots, it can be done through such a requirement. But if voters don't want to rank everyone, allowing them to truncate their ballot is reasonable policy decision. It also has been seen as a reasonable decision to give voters three times the chance to cast an effective vote with IRV as they would have with a plurality ballot, although most IRV backers would prefer a ballot with unlimited rankings (as indeed will be what San Franciso uses when it has voting equipment that allows it). —Preceding unsigned comment added by RRichie ( talk • contribs) 14:35, 27 January 2008 (UTC)
User:Tomruen moved the detailed election information to a new page he created. There are now *three* pages on the history of IRV, and none of them really address much of the history! In the process, without comment, he removed the election analysis that summarized results from IRV elections so far.
I think there is a basis to have subarticles. One, a history of preferential voting (using the STV model). This is general, not confined to the United States. It would have a section on the U.S., a brief summary. Second, a history for the U.S., so the two pages that now exist should be merged. The other one on the U.S. was created as another POV fork; the creator, a probable sock, simply put back in what had been taken out of this article based on improper sourcing or for other reasons.
The preferential voting/STV article currently is called History and use of instant-runoff voting. This should use a name by which the method has been known for almost its entire history, all but the last decade, rather than the newly-coined instant runoff.
The other U.S. history article is History and use of instant-runoff voting in the United States. And now, of course, Ruen's IRV implementations in United States.
However, each article should stand alone. If there is a history in the U.S. article, it should be summarized here, using WP:SS, and the summary should be NPOV. Among other things, that means that there should be balance. A list of adoptions, which is what Ruen left here, is effectively POV. It's promotion. Most of those adoptions are moot, some may never see an actual ballot.
The forks or subarticles should have names that have Instant-runoff voting at the beginning, to follow the convention that puts the articles together in alphabetical indexes. That's what we did with Instant-runoff voting controversies.
We started out with a list of adoptions in the U.S. I added detail, such as describing which ones had actually been implemented or which were actually scheduled for elections, plus a brief summary, when the information was available, of how IRV was performing in terms of its goals. IRV is being implemented in jurisdictions where, previously, there were top-two runoffs. So how it performs in comparison to top-two runoff is highly significant. But, I can imagine, it was not liked that this information was being noticed, so ... with Ruen's move, it was taken out. Nice work, Tom. -- Abd ( talk) 05:01, 27 January 2008 (UTC)
Abd - you added selective detail. If you're going to write about the specifics of IRV elections, it makes sense to be much more comprehensive than these articles have been. Putting it in a separate article that is flagged as incomplete makes sense, doesn't it, rather than cherrypick examples? —Preceding
unsigned comment added by
RRichie (
talk •
contribs) 14:37, 27 January 2008 (UTC)
Still learning Wikipedia here - I'll start signing these.
Also, Abd, I had a concern about what you're doing with these edits on San Francisco, etc. Aren't you doing "original research"? Wikipedia says that A"aticle policies" include "no original research", but that's exactly what you're doing. Your mistakes are exactly why I assume original research is frowned on. I think you should put this kind of information on your own website (Center for Range Voting, which I know you post things too). RRichie ( talk) 16:57, 27 January 2008 (UTC)
I moved two talk sections to Talk:IRV implementations in United States HOPING this will solve the unnecessary expansion of details in the main article. I also expanded the bullet-list format there to sections so each city can have extended information as details and results progress. Please move all discussion there. Tom Ruen ( talk) 07:10, 28 January 2008 (UTC)
I just removed the {{ POV}} tag from Approval voting, because all of the concerns in the section discussing it have apparently been addressed.
For here, I'm not so sure. I know that we have had some serious disagreements, but things look to have settled over the past week pretty well, at least as far as the neutrality concerns that I recall are involved. It's hard to tell because all of the POV tag discussions seem to be archived.
Are there remaining neutrality issues, and if so, what are they all? MilesAgain ( talk) 02:16, 28 January 2008 (UTC)
Section talk moved to: Talk:IRV implementations in United States Tom Ruen ( talk) 07:03, 28 January 2008 (UTC)
I removed this paragraph from implications section:
I didn't like the soft some other voting methods (What methods!? Bucklin or Borda?). I didn't like the soft could prolong the counting procedure (compared to what?) The entire paragraph looks to me like whining. I'm content if something is restored that is more objective, perhaps:
BUT basically the two examples in the final paragraph already said it better!
I know IRV-opponents want to use this summability limitation as a reason to discount IRV, and happily claim whatever alternative rank method Borda count or Bucklin voting or Condorcet method or whatever doesn't have this limitation. BUT you can't just say A is bad because there's a small cost, while the cost to alternatives is neglected (i.e. Voters don't WANT lower choices measured AT ALL unless their choice is eliminated, SO it's the nature of runoffs themselves.)
I can't really support ANY comparisons to other methods under implications, since implications ought to refer to changes from what IS, not alternatives which do something completely different.
Anyway, I figured this was as good of a fight as I'm willing to offer for now. Tom Ruen ( talk) 02:36, 30 January 2008 (UTC)
P.S. The Summability criterion was one of the articles deleted from Wikipedia, and reconstructed on Abd's user subpage: User:Abd/Summability criterion.
P.S. If Electoral fraud is an issue with sources, perhaps it ought to be included. I accept compared to plurality, it's a legitimate issue to consider. I mean it seems much bigger - like all votes should have a paper trail, and ballots should be designed in a way that allows fraud to be detected. For example, a checksum could to be performed at each precinct, counting how many marks there are at each ranking (like Bucklin sort of), so if marks are added (or removed?) after the voter is finished, it can be detected. I don't know ANY sources for such concerns, but it seems obvious to me that people in-the-know must think about these issues. Tom Ruen ( talk) 03:00, 30 January 2008 (UTC)
User:Peter Ballard added the following text (in my italics) to the article:
Now, the problem is that this text draws a conclusion. First of all, the facts are going to need a source. However, the "facts" don't make the conclusion. Somebody must. Who? It's fairly clear that the National Party exists, and has for a long time. However, if I'm correct, it's dying out. It used to be a more or less equal partner with the Liberal Party. If it is dying, this is *confirming* similarity with plurality, rather than being a "significant difference." IRV, pretty clearly, allows far more flexibility in voting, *theoretically* allowing third parties to function without spoiling elections due to vote splitting, but it should be noted, if third parties run on a par with the two major parties, a nastier effect can appear than the simple spoiler effect, there is "center squeeze," a phenomenon which is rare when there are only two dominant parties, and I think this is behind Duverger's law. (When Center Squeeze is taking place, IRV is discarding the candidate actually favored by a majority over either of the other two major candidates, but with fewer first preferences, which is essentially behaving like plurality.) There are other methods of Preferential voting that might not have this problem. The *real* problem is, of course single-winner, which is, in my view, very inappropriate for representation, even if the method works and the majority votes for the winner, it leaves a minority, sometimes almost half the electorate, unrepresented. That's seriously defective. "No taxation without representation" was a founding slogan for the United States. Okay, who represents me? Am I represented because someone else appoints my representative? If so, what were the American colonists complaining about? They had representatives appointed by the King!
(To be sure, "representation" was thought of in those days as being representation of communities, not of people. But, again, it was the people who were taxed, not, generally, communities.)
If the survival of the National Party is to be mentioned, *how* it is surviving, in what numbers, with trend over time, would need to be mentioned as well. People, in the U.S., are interested in these things, sometimes, because they are supporters of third parties, and would like to know what the ultimate effect is. If the ultimate effect is that the party dies a slow death, that's quite different than "thriving," and isn't even fully "surviving." Dying takes time. Is there any example of a third party *arising* under IRV, and growing to become a major party? *That* would be interesting, absolutely. -- Abd ( talk) 17:25, 30 January 2008 (UTC)
Is it really controversial to suggest existence is aided by a lack of need for vote-splitting from a third party supporters and a similar major party competitor? Tom Ruen ( talk) 21:42, 30 January 2008 (UTC)
-- Abd ( talk) 22:02, 30 January 2008 (UTC)
Removed:
This statement is without proper context since the only comparisons that will effect parties and candidates are plurality or two round system which have the same behavior! Tom Ruen ( talk) 22:36, 30 January 2008 (UTC)
IRV elects candidates who have a majority of votes among those voters who express an opinion between the two finalists. Abd repeatedly seeks to insert the point that IRV winners may have less than 50% of ballots that had a vote for a candidate in the first round. Whether such candidates are "majority winners" is important to his POV. This distinction appears at various points in the article, and probably should be reduced to one statement early in the article so as to avoid having to repeat the concept over and over. For example, Abd has inserted the unique example of a Vermont bill in which exhausted ballots can result in "no election" by the voters. I know about this in detail, because I wrote the bill. This language was crafted to fit with the language in the Vermont constitution, which says there is "no election" if no candidate has "the major part of the votes." If there is "no election" then the constitution charges the legislature with selecting a governor. While that language is not crystal clear, so that exhuasted ballots might still be discounted, there is a principle of constitutional law that legislation that makes a constitutonal provision meaningless is unconstitutional. Therefore, to avoid making the supplementary role of the legislature in selecting a governor a theoretical impossibility, the bill was crafted in this unusual way.
I cleaned up Abd's sentence to give it a citation and explain just a bit...However, I think the sentence Abd seeks to include citing the rare case where exhausted ballots can result in "no election" is so unique, and only exists as a bill and has never been implemented anywhere in the world, that we know of, means it should not appear in the Wikipedia article since it distorts the general understanding. Government elections are assumed to be decisive, with the option of None Of The Above (or exhausted ballots) leaving an office vacant, or resulting in a new election, existing only in theory. I have not deleted Abd's insertion about this unique proposal in Vermont yet, to give others a chance to look it over first. Tbouricius ( talk) 15:41, 31 January 2008 (UTC)
User:Tbouricius edited (properly) my mention of the Vermont situation with IRV, with the summary, "(Changed Implementation" to "proposal". and added cite. However, I think this whole sentence should properly be deleted as a rare exception.)"
Here is why it is actually very important. IRV is (almost?) exclusively being implemented in jurisdictions which have majority election requirements, and therefore, sometimes, runoff elections, which are allegedly expensive, providing an economic incentive for implementing preferential voting, such as IRV -- (and only recently has there been much agitation for any other form or runoff-avoiding reform). The most notable, of course, being San Francisco. Jurisdictions have those requirements because, apparently, it's believed that a majority of those voting should consent to an election. This is, as Bouricius knows, a common belief, supported by Robert's Rules. In San Francisco, IRV was sold with an assurance that "a majority would still be required." Yet, in fact, a majority of those voting is *not* required. The ballot summary was misleading. Voters show up, vote for an eligible candidate, and yet their vote is disregarded when it comes time to determine a "majority." That is a blatant violation of the underlying principle.
It is often argued that, well, if voters don't show up and vote in a real runoff, their votes don't count also! And it's implied, as this article has implied from time to time, that exhausted ballots are the voter's fault or choice. Perhaps, sometimes, but perhaps there were a lot of candidates (more than three) and the voters used all the available ranks by voting sincerely. Or some didn't understand the method, or really detested all the remaining candidates. The fact is that, in the large majority of "instant runoffs" we have seen since San Francisco took the plunge, a majority of voters showed up and voted for candidates other than the winner, not for the winner. If a runoff were actually held, we may suspect, roughly one-third of the time the election would reverse from the plurality winner in the first round. (examination of real runoffs, which will show similar voting patterns in the primary, shows this or something similar, plus we have some analysis on this point from Australia). Some of that may be due to differential turnout, but ... quite likely not all of it (and differential turnout doesn't apply to the Australian data, of course).)
(And if that last-round definition was the meaning intended by "majority," it was thoroughly deceptive, for there is no "requirement," IRV defines the winner as having a majority simply by discarding all votes not for the top two. As I've pointed out, we might as well take it one more round and call it "unanimity." It's not a *requirement*, it is a tautology.)
IRV, as implemented in San Francisco, has defeated the purpose of requiring a majority. By passing the proposition, voters, quite likely without realizing it, discarded the requirement, substituting something else that was sold, through the use of the politically designed name, "instant runoff voting," as being equivalent.
Now, I'm sure that Vermont legislators and IRV activists would do the same thing if they could. What is unusual about Vermont is that the majority requirement is in the state constitution. They can't simply legislate it away, sweeping the issue under the carpet, as was done in San Francisco. So, yes, it is rare, but only because a constitutional requirement for a majority is rare. Many jurisdictions have that requirement, so the idea of requiring a majority isn't rare.
Apparently, this argument has mostly escaped the opponents of IRV. Nothing like this was brought up in all the ballot arguments provided in the San Francisco proposition, and, to my knowledge, it hasn't been raised elsewhere in real campaigns yet. I'm sure that is about to change. This argument, and the data behind it, is now becoming known to activists working on the issue, I don't think it will continue to be possible to sweep this under the carpet. An ounce of experience is worth a pound of theory.
In any case, I'm not suggesting that this argument be put into the article. However, the underlying facts exist and can be properly sourced. They were put into the article, and removed by User:Tomruen as part of his creation of the U.S. history fork. I'll be bringing them back when I have time. The summary of how IRV is performing belongs in *this* article, and, if anyone claims it is cherry-picked, then they are welcome to add the allegedly excluded facts. Such as, for example, if anyone wants to do the searching, how IRV has affected turnout. Certainly that would be relevant and interesting. As well, how has IRV affected election costs, overall. Problem is, it's hard to get that data. Still, if someone can find facts relevant to it, they should be in the history article in some detail, and in this article in summary form. -- Abd ( talk) 20:36, 31 January 2008 (UTC)
I added a test table on the largest IRV election from SF, 2004, could use some comments - Is it useful to include in the IRV_implementations_in_United_States article? Tom Ruen ( talk) 02:21, 1 February 2008 (UTC) Talk:IRV_implementations_in_United_States#San_Fran_results
I think so. I'll comment there. Nice work. Tom, you will learn more if you look at the Excel spreadsheets they provide. These give the full ballot data, summed, i.e., we know how many voters voted for each candidate in each rank. If this had been a Bucklin election, more of the candidates would have reached an absolute majority, and, of course, we'd have all the relevant data. As it happens, Bucklin would have given the same results as IRV in all these elections. (Which isn't surprising. The differences arise in unusual cases.) -- Abd ( talk) 04:10, 1 February 2008 (UTC)
User:RRichie removed the following italicized text from the section on practical implications: diff
The statement preceding it is not true under the conditions that a majority is required. Every jurisdiction actually implementing IRV (except I have not checked out Takoma Park, I assume Mr. Richie could tell us about that one, I presume he sold them the fish bicycle) did so to replace majority-required elections, and IRV was promoted as a way to get majorities without holding a presumably expensive or inconvenient by-election. Indeed, it can do this, but it can also fail, and, it turns out, it is *usually* failing. The reason why a majority is not required in the current IRV elections is that IRV *replaced* the majority requirement, it did not satisfy it, but this appears to have escaped the notice of those who might have been opposed. The voter information pamphlet for the San Francisco proposition did not even mention the argument (among all those who paid for opposition arguments); and to the contrary, the summary language, supposedly neutral, explicitly claimed that candidates would "still be required to gain a majority of votes." (I think that's an exact quote, but it's from memory, might have been a little different. I quoted the exact language in Talk; in a section that was moved to the Implementations article). "Majority of votes" has a common understanding and meaning: take all ballots containing a vote for an eligible candidate and count them. Divide that number by two. The winner must gain more votes than that. The "majority" in IRV is a majority of a reduced subset.
I consider it highly likely that, in the future, new IRV initiatives in such jurisdictions, in order to pass, will retain the majority requirement, for the cat is out of the bag. In my view, IRV with a majority requirement is substantially better than IRV without one. Without the requirement, IRV creates a "majority" by discarding all votes from consideration that don't rank the top two; this results, necessarily, in a majority *of votes remaining*. Take any election, with any method I can think of, and discard all votes not for the top two. Presto! A majority for the winner. Indeed, if IRV can be claimed to guarantee a majority winner, we might as well claim that it always finds a consensus winner, elected by unanimous vote. Just continue the elimination process one more step, until there is only one candidate instead of two! No, IRV does help find majority winners, that is true under some conditions. But it can also miss them, because of the quirks of sequential elimination, as I'm sure Mr. Richie knows quite well. Indeed, he sometimes argues that this is a feature, not a bug. IRV requires candidates to have a certain minimum "core support." Bottom line, this means that the winner, generally, must be one of the top two in first preferences. (With batch elimination, used with some elections that Richie cheerfully calls "IRV," that is exactly it). Apparently, this is more important than having winners actually approved by a true majority of those who bothered to show up and cast a valid vote. I.e., standard majority rule.
Since we are on it, there is one standard response to this line of thinking: but runoff elections are often decided by fewer voters, with the winner getting fewer votes than loser got in the primary election. First of all, this probably assumes relative inconvenience for voting, thus lower turnout, in the runoff. In Cary, the reverse was sometimes true, and generally runoff elections enjoyed about the same turnout. But, it is argued, by compressing all voting into one session, it is more convenient for voters, and therefore IRV is more democratic. This is actually a far more complex debate than one might think. Runoff elections probably choose more satisfactory winners, overall, due to differential turnout. Essentially, runoffs test preference strength, the voters who don't turn out for the runoff *don't care as much about the results* as do those who turn out -- generally. Obviously, there are plenty of individual exceptions, such as parents who can't find child care, etc. But these considerations can apply to both elections, and runoffs allow voters who *do* care an additional opportunity to vote. As anyone familiar with Range voting would recognize, this is a true test of sincere preference strength, on average, thus votes are, as with Range, deweighted according to such strength, and it's easy to show, mathematically, that this should result in higher average satisfaction with the result. If the top two are equally acceptable to me, as an example, I may not bother to vote, and, in fact, if that's the only issue, probably won't. So turnout can be lower, and possibly a lot lower, even if there is no differential barrier to voting.
Point: many elections do require a majority for victory. Some of these have election/runoff schedules that may actually favor the runoffs. (Think about the current overall Presidential primary process as the "primary," with the top two facing each other in the general election. This is *roughly* what we have. When IRV replaces a majority-required runoff, depending on how the replacement bylaw is worded, majority might be continued to be required (the Vermont legislation proposed), or not (the San Francisco proposition as actually worded, not as advertised).
Now, as to balance. Richie's edit summary was: "under "practial implications", doesn't seem useful to describe in laborious detail a proposal that no governmental election uses."
This article, first of all, is about voting systems, and not only about those used for governmental purposes. Secondly, the detail given is very brief. Such legislation is pending and, indeed, should be covered by the article (or at least the implementation article).
We have seen this "laborious detail" argument many times. I will refrain from bringing in all the political implications, what I know about Richie's arguments elsewhere and how this dovetails with them, but it is important to note here that, obviously, User:RRichie is Rob Richie, Executive Director of FairVote, the main advocate of IRV. He was here before as a tendentious IP editor, actively removing contributions from "critics," -- yes, that was the edit summary, "editor is a critic" -- and was blocked as such. He is quite welcome to participate here, as far as I am concerned, for he has great and even unique knowledge of the subject. For example, we have a report as to how the term "instant runoff voting" came to be used, but it would certainly be of interest for the article what Richie recalls of that. (Never mind for the moment how we would find reliable source, suffice it to say that there would be a way.) However, he is a COI editor, even more so than User:Tbouricius and, as such, should refrain from any sort of editing that could reasonably be seen as promoting a point of view where he has a conflict.
The judgment of what is "useful" depends highly upon POV. Useful for what? Richie is a political activist, making his living at influencing opinion. Any skillful activist knows that what is not said can be as important as what is said. If one can place before decision-makers a subset of information, chosen to further only one point of view, one can influence them toward that point of view. Other information certainly won't be seen as useful! What is important for readers to know? The fact that IRV disregards, as commonly implemented, inconvenient votes looks bad. But if we can blame it on the voters, by how we state the fact, i.e., with language like "voters who did not bother to rank enough candidates" or the equivalent, i.e., "failed to rank" has been used in the article (and might still be there), one can divert attention from the fact that an election was held, a majority of voters voted for someone other than the IRV winner and did not vote for the IRV winner, but the IRV winner still wins. One may certainly then argue that for various reasons the IRV winner was still legitimate, but these same arguments start to resemble the arguments for plurality. Why do jurisdictions bother to require a majority? Are they aware that they are waiving the requirement if they implement IRV as generally proposed? In Vermont, quite simply, the requirement could not be waived, it is constitutional. *That is the only difference.* This is actually a *crucial* fact. Now, does this have POV implications? Certainly! That is, there is a clear POV motive that one might have for inserting this into the article.
"It will confuse the readers." Sure it will. Without it, they will be quite clear, look at this, all the evidence supports it, etc. Add some more facts, they become "confused." I.e., realize, more, the complexities of the situation and no longer find it so obvious. And out of this confusion can come true clarity. Wikipedia policy trusts the readers to sift through the evidence, but if the evidence is differentially presented, the results can be warped. Wikipedia actually depends on editors with POVs, they do, in fact, most of the work. Someone without a POV on IRV probably hasn't been paying attention! If one has a POV, one can become very sensitive to opposing POV, that might otherwise pass unnoticed. For easily more than a year, I assumed that FairVote was right: "Roberts Rules recommends IRV." It was only my POV that caused me to look more closely, why would this very respected source for parliamentary procedure make such a mistake? Turns out, they didn't make a mistake. RR doesn't "recommend" IRV -- and, of course, this has filled the Wikipedia drives with many kilo-characters of discussion which do not need to be repeated. And, again and again, pro-IRV editors would take out the necessary detail, put in to make the mention of IRV in Robert's Rules accurate and neutrally presented, not excluding contrary detail.
Because of the COI, Mr. Richie should refrain from making edits which could reasonably be expected to be controversial. If he wants to make a controversial edit, he should instead, discuss it here to seek consensus, and not be the one to actually make the edit (unless no controversy actually appears). For details and confirmation of this, see WP:COI. And, in spite of all this, once again, welcome, Mr. Richie. I assume we can cooperate toward the improvement of what has become a family of articles, and my goal has been to help raise this article to Featured Article status. That will require, however, thoroughly cleaning up any remaining POV imbalance (in any direction). As Mr. Richie knows, I generally oppose IRV as an inferior method, but it is quite reasonable for some jurisdictions to try it, and my opposition was originally based almost exclusively on "it is a good reform, but we can do better at lower cost." As I found with some of those who came before me (i.e., people who have been studying elections methods since before IRV was named), I've actually come to oppose it for various reasons, mostly that it doesn't bring sufficient improvement, in practice, to be worth the cost, or, at least, my support has become much, much weaker, more easily influenced by cost. It does not escape my notice, though, that the 2000 election would have had a more democratic outcome if, say, IRV had been in place in Florida for presidential elections. But the same is true for the much, much simpler reform of Approval voting, cost-free, or the alternate "instant runoff" of Bucklin voting, which has the advantage of clearly satisfying the majority criterion, but which is also more efficient at finding majority winners -- it does not miss any votes from the counted rounds. (I need to put the info back in, it looks like I failed to save an edit, but, as I recall, Bucklin analysis applied to the San Francisco election data did find a majority winner in almost all the elections where IRV failed to do so. Bucklin, which requires no changes to equipment or significant change to procedures (Just Count the Votes), really deserves more attention.) -- Abd ( talk) 18:35, 1 February 2008 (UTC)
(unindent)First of all, my thanks to Mr. Richie for taking the time to copy the material from the Encyclopedia Brittanica. It will be interesting to examine it from the perspective of this problem. As to his having more knowledge than I about instant runoff voting, of course he does. He's been promoting it for more than a decade and has created many of the arguments advanced for it. He is not, however, an expert on Voting systems in general, that is very clear from much of what he has written. His vision is narrow. Nevertheless, it is very welcome here. Contrary to what it sometimes seems, expertise is not a disqualification from editing Wikipedia. But experts often don't have the patience necessary. It's a problem, actually.
However, as to Wikipedia process, I could properly be his guide, even though it can be truly arcane and unpredictable. If we can find consensus here, we'll quite likely end up with a better article than if we try to take issues to mediation or arbitration. Now, Wikipedia does *not* expect editors not to make mistakes. We can go out and find statistics and report them. If we find a mass of statistics, we may very well make mistakes. In fact, I expect it for myself. Wikipedia is a collaborative effort, far more than traditional publishing. If I had submitted those San Francisco statistics for formal publication, you can be sure they would have been checked and double-checked before the manuscript went off. Here, I wrote the Talk page edit as I collected the data from the source given, so anyone could check it. In this case, there was an interpretation error, but I'm sure it would have been caught in any normal editing review. None of this should change that was reported was still *substantially* correct. IRV is not performing as I would have expected, nor as it was promoted. It is avoiding runoffs, but the same results would have obtained, in all 33 elections, if a majority requirement had simply and directly been discarded, as it was discarded by implementing IRV, but instead going to ordinary Plurality voting(or, better, Approval voting, or, even better from some points of view, Bucklin voting. The same result would have been obtained if all the elections had been Bucklin, except that all but one election would actually have found a majority -- if we assume that voters had voted the same ballots, and there really is little reason to expect that voters would have added fewer second and third rank votes, Richie's argument -- that I'd expect him to make -- on that is weak. Voters aren't going to consider, most of them, later-no-harm, which is of rare consequence in real elections.
IRV failed to find a "majority of the votes" in seven out of nine elections going to instant runoff, as I recall the statistic. One of the interesting aspects of Brown v. Smallwood is that the court explicitly -- and correctly, based on precedent -- associated votes with "men," not with marks on paper. (Times have changed!). Thus votes really means a majority of voters participating in an election.
Mr. Richie has misrepresented my arguments about IRV and majority. IRV with a majority requirement is a different than IRV without it. I would argue, in fact, that it's a better system with a majority requirement, more democratic. IRV does *not*, in one round, reliably find a majority winner, under all legal definitions of majority, including those of Robert's Rules. Implementing single-ballot preferential voting with IRV or any other preferential method does *not* guarantee a majority, unless you force voters to rank all candidates, which would go over in the U.S. like a lead balloon. So implementing IRV by legislation superseding majority-required legislation is *removing* the majority requirement, not merely finding a cheaper way to satisfy it.
(But I would have thought it would be more successful at satisfying it, I *do* consider that if a majority of votes, after transfers but still counting all the votes as the basis for majority, are for a candidate, this is satisfying the majority requirement. I don't like some aspects of it, the candidate eliminations, which is why I'd favor other ways of analyzing a preferential ballot, but the fact remains that, by voluntarily ranking candidates, the voter has voted for them, and thus may count as having voted for the winner. Coercing the voters into ranking all candidates is not obtaining the free consent of the voters to a result. Bad idea.)
There is nothing in the Brittanica quotation that contradicts this; on the contrary, the Brittanica article confirms the point. One might notice this: "This redistributive process is repeated until one candidate has collected a majority of the votes." This is in reference to Australia, where compulsory full ranking causes the process to succeed in this. There is a difference between "majority of the votes," which language is present in the Vermont constitution, I think, and in any case which refers to all valid votes. Votes which have been substituted are valid votes, and remain such. Exhausted ballots do not thereby become non-votes. They are still votes, simply for none of the remaining candidates. With the Mayor of London, if you didn't vote for the top two, in your main vote and your supplementary vote, you are simply out of luck. Your vote is disregarded, and a "majority" is cobbed together only by discarding your vote quite the same as if you had cast a blank ballot. But you did not cast a blank ballot, and perhaps you even voted for a candidate on the ballot. "Majority of the vote" is a concept which has been used for centuries in deliberative process, and if we are going to discard it, it should be deliberately done, not covered up by pretending that we get a "majority of the vote" by discarding some of the legitimate votes.
Note, however, that in partisan elections, it is quite likely that most voters under suppementary vote would, indeed, vote for one of the major candidates in first or second preference, so this system would almost always find a majority, only in very close elections would it not, sometimes.
It's simple, here. If the fact can be verified, and it is notable, it belongs in the encyclopedia, somewhere, and something directly contradicting it can't be maintained, anywhere. (But if there is controversy, then the sides may be reported as such, and then relative notability can become an issue.) It may even be non-notable, but if it can be verified, a clear contradiction cannot stand. If some source adds 2 objects and objects 2 and reports a total of 5, and it happens there is no other reliable source on the sum of the objects being counted, we don't put "2 of these and 2 more makes 5." We fix the error (and we might note it as such). When sources write "majority," in regard to IRV, they are often simply talking about "majority of the votes remaining after exhausted ballots are set aside." Note that it has been a struggle to even get this into the article, what with the barrage of sock puppets and POV edits.
This article, I must remind Mr. Richie, is not just about IRV in public governmental elections. It is about the *method*, which is proposed for use, sometimes, in organizational elections, the sum of which probably exceed public elections by some measure. (But not necessarily in terms of numbers participating each year.) Further, various organizations, including governments and political activists, are considering reforms, and thus need to understand the *possibilities*, not merely what already is in common use. Wikipedia rules restrict us from putting much here about what is well-known about election methods, because much of it is known from peer discussion, outside of the formal literature. Further, even the formal literature is only accessible to a few.
To bring this back to the point, IRV, when used to replace top-two runoff, not only may fail to find a majority winner, it seems that, in the U.S., it usually does, from the examples before us. (That is, if the election fails to find a majority in the first round, in almost every case, it still fails after vote transfers are completed.) I'm sure this would be of interest to those considering IRV for such replacement. Is it important that a majority of those voting have approved the winner? If so, IRV isn't a full solution. Indeed, no single-ballot election method is a full solution, covering all possibilities, unless we coerce voters into ranking all candidates. Further, it is troubling that a majority of voters -- actually as many as two-thirds of the voters in a totally extreme case -- can vote *against* a candidate in favor of another, and yet that supermajority-rejected candidate can win with IRV. Unlikely to be that extreme, yes, very. But the possibility of a majority winner existing who is passed over by the IRV process is much higher. How high is hard to predict, because of the lack of actual ballot data, but center squeeze is quite reasonable under some conditions, where there is a polarized electorate balanced between two candidates and a good compromise candidate, second choice of most people, first choice of less than both of the other two. As Richie knows, there are other preferential voting methods which don't have this problem, including one which is much, much simpler to count, Bucklin.
From the San Francisco data, if the goal is to replace runoffs, Bucklin did find a majority winner -- the same as the IRV winner -- in every contest in 2004 except the very hard case of having 22 candidates (it was just a tad shy). And Bucklin is very, very easy to count. It is also, it can be claimed, a form of "instant runoff voting," but using a different means of estimating who would win the runoff. In most cases the result will be the same, but not in a center squeeze situation, where the compromise winner will likely prevail. In the U.S., we had STV for proportional representation (good method for that), and Bucklin for single-winner, in a number of places, before the backlash took out all but Cambridge. Definitely, Bucklin should be on the table. But much of this is dicta, here. It is mentioned to show the importance of the issue of "majority of the votes." Which is distinct from "majority of the remaining votes. -- Abd ( talk) 22:10, 3 February 2008 (UTC)
Can we add a section under this name and include differences causes by this requirement and stop with the detailed exceptions all over that make the article hard to read and understand? Tom Ruen ( talk) 21:30, 1 February 2008 (UTC)
Well, folks, for a few months I've been noting that MilesAgain was a sock puppet. It's important because even legitimate sock puppets should refrain from contentious editing. I had suspected he was a sock in the same series as User:BenB4 and User:Acct4 who had been part of the cabal of editors "owning" this article; however, he had been more conciliatory, more willing to compromise, and his disruptiveness stopped just short of motivating me to file a WP:Request for checkuser. Well, now we know why he stopped short. I'd have filed it as a suspected User:Nrcprm2026 (James Salsman) sock, as were the others, it would have taken him out, almost certainly, quite a while ago. He's been tagged and blocked entirely without any action on my part. Thought you should know.
Just on a hunch, I googled "James Salsman" "instant runoff voting." Came up with this:"Though we don't even have a Chapter in Santa Clara County, some of our active members there have got IRV on the ballot, as you know. Kudos to Steve Chessin, James Stauffer, James Salsman, and Russ Paielli! (Hope I didn't leave anyone out!) [1]. Distinguished company. This was from 1998, and I recognize three out of four of those names, and I'm sure some of our other editors will as well.
Here is a URL for an open Salsman post on this topic: [2]
Unless it turns out that two checkusers erred in the identification of MilesAgain as another Salsman sock, he lied to us, explicitly.
-- Abd ( talk) 18:47, 3 February 2008 (UTC)
What happens if the two candidates with the smallest number of votes have the same number? Who gets eliminated? Timeshift ( talk) 23:06, 3 February 2008 (UTC)
I moved content before January 23 to a new archive 5, so my scrollbar will function reasonably again, maybe for another 7 days with luck? Tom Ruen ( talk) 23:58, 3 February 2008 (UTC) Talk:Instant-runoff_voting/archive5
Surveying sentences refering majority in the article (by section in bold), here's what I found:
Intro:
Tom: Majority definition okay on "first round". I'd support a change the second part to a logical description:
Abd:need to have the qualification, if it is not already plain from context. Given the context, I fixed the small error in Tom's language (normally that's not a good thing to do in Talk, but I expect he won't mind in this case.)
Tom: I don't think this belongs in the intro, even if its interesting. I'd merge statement into Non-governmental organizations section with RNRO.
Abd: It's been taken out. Note that this is one of the options I proposed long ago (and implemented). However, certain IRV advocates, one of whom we now know to have been James Salsman, insisted it belonged in the introduction, but attempted to get necessary qualification or context removed. It was very clear POV pushing. My "political agenda" would actually favor the complete explanation to be in the introduction ... but I am *not* editing this article to pursue a political agenda beyond one served by a fair and neutral article on the subject.
Counting the votes:
Tom: Seems qualified and clear
Abd: Maybe. I think we should use uniform terminology. "Active" is a synonym, and undefined, though a reader might reasonably infer the meaning. We should also use the language used in legislation and reliable sources, if possible, and not invent synonyms unless there is utterly no doubt about them. I certainly considered "active" close enough not to edit war over! But i don't think it optimal.
Example:
Tom: Properly qualified, but I think its over-kill to focus so much on this fact with a single example.
Abd: From what has become plain, "majority of total ballots" isn't the language to use. It's "majority of the votes," and we can make clear what that means early in the article so that we don't have to keep explaining "majority of the number of ballots containing at least one vote for an eligible candidate," or some other definition probably taken from a legal source (or Robert's Rules), whence it may be sourced.
Effect on parties and candidates:
Tom: This statement is unclear meaning to me - what does absolute majority mean? I can guess, but I'm not even sure this fact is important enough to give here under a Effect on parties and candidates section. Abd: It's an error, I think. But Australian usage may be different. So this may take some thought. I think it means "absolute majority of the voters participating in the election, having cast a valid ballot." That defines a group membership without specifying, beyond the fact of valid participation, how they voted. Because this is Australia, though, "absolute majority" is guaranteed, it's a tautology, there are no exhausted ballots because if a voter truncates, it's not a valid vote. Am I right about this? So this was a POV edit, though it may not have been intended that way. It was an Australian not understanding the context of the debate that was going on in Talk, I'd guess.
Effect on parties and candidates:
Tom: I think I understand this even if ambiguous. It's related to the power of gerrymandering. Your party can get a minority of the votes and a majority of the seats if you can pack the opposite party supporters more densely into a few districts that win by landslides so other districts can lean towards closer elections in favor of your party.
Abd:I'm not sure this should even be in the article. The intention is perhaps what we might call an "IRV majority." If it meant "majority of the voters participating, casting uncoerced votes," then IRV doesn't guarantee that even locally, not to mention nationally. Nor does any other method. (Methods cannot create support! Methods can only discover it if it exists.) This is really a "single-winner" problem, not an IRV problem. If we could get this IRV nonsense out of the way, we could really cooperate on proportional representation, though we might still debate about which method to use. Frankly, though, STV is right up there among the top choices, and is one with a lot of experience.
Majoritarianism and consensus:
Tom: This whole section seems an unrelated mixture of ideas - talking about majority vs PR on one side, theoretical Condorcet comparisons on the other, and explains the definitions of majority.
Tom: I have to think whatever is worth saving here, it needs to be split into each issue:
Abd: if Condorcet ballots may be truncated, or if they otherwise indicate some distinction between Approval and non-Approval, then it is possible to determine if there is a majority or not. Otherwise, what we would have to say is that a majority of those indicating a preference between this candidate and all other candidates preferred the Condorcet winner. There are no "exhausted ballots" in Condorcet. However, there may be ballots that do not participate in a particular pairwise election. Question is, did a majority indicate a preference for that candidate over all others? That can be determined. Condorcet does not require a majority, but Condorcet methods may be used in elections that do require a majority.
Is IRV better than other systems?:
Tom: The first two go together as and come down to differing definitions of majority. The THIRD is something completely different that I'm sure will confuse people, referencing the Condorcet winner without saying this.
Abd: This section is an arguments section, and arguments are presented as they actually are presented in the real world. These should really be quoted and, no, we don't get to change them to make them truer or more false. The "majority" in the last mention is "majority of the votes cast," or, more accurately, of the voters voting and casting valid votes (i.e., votes that would have been counted if for a winner).
Similar systems:
Tom: NOW apparently we have a NEW definition of majority (approval majority?!) which allows multiple votes, multiple majority candidates, but only one winner. This is NOT good as is in my view!
Abd: Its the same definition. A majority of those voting have, in that round, voted for the winner. Yes, with Bucklin, we have to realize that is is possible (though actually quite unlikely, I haven't seen a real Bucklin election which did it) for more than one candidate to gain a majority. The "lower choice candidate can count against your higher candidate" is a POV summary of Bucklin, highly imbalanced. Show an election where this actually happened and the voter would have been upset about it! The definition should be rewritten, it does not actually explain Bucklin counting (but it might continue to). Again, since it is in the negative, the double majority problem *does not exist*. The winner is, as with Approval, is the candidate with the most votes when counting stops, and it stops when a majority is found. That there might be two or twenty is irrelevant. One stops the counting, the winner is the one with the most votes.
Runoff voting:
Tom: Overall majority?!
Abd: Synonym for "majority of votes cast. Totally unnecessary, since there are no eliminations in the first round. If there is no majority in the first round, there is a new election with reduced access to the ballot. Often, in theory, that new election is still open and majority failure could occur, but I think that this is very, very rare.
IRV in a larger runoff process:
Tom: I don't know where to start, but maybe this summary is helpful. Tom Ruen ( talk) 23:36, 3 February 2008 (UTC)
A difficulty in terminology in this article is the use of the word majority can be defined differently:
No election system can guarantee a majority of voters will support a single winner and the only way to make a promise on majority is to redefine it in a way that exclude voters from the count who refuse to offer a vote among the final set of choices.
If an election implementation requires a majority winner, then it must be defined in a way that limits a winner from being declared and leaves the election result open to a further election process.
Some elections, including variations in IRV implementations, may allow voters a chance to offer a protest choice like None of the above that rejects all choices offered which may binding or nonbinding. If nonbinding it has no effect on the election. If it is binding this vote can block a winner from being chosen.
First of all, majority has a clear definition: "more than one-half." What gets unclear is when the word is used without specifying "half of what"? absolute majority, according to the article, means "half of all eligible voters," not merely those who voted, but including those who did not vote. Absolute majorities are required for certain actions (for example, an absolute majority generally may be able to amend bylaws without notice; because absolute majorities can be hard to obtain due to low participation in many organizations, supermajority rules were developed to make it practical to amend bylaws without jumping through such a high hoop: after proper notice, a 2/3 vote may amend bylaws.) However, the term may be used, and I think I've seen it used that way, in an election as "absolute majority of the votes cast;" the "votes cast" being the number of ballots containing a vote for an eligible candidate for the office. IRV may proceed to ignore some ballots with such votes, but, then, it is finding, not a majority of votes, but of a subset of votes, being those remaining after eliminations. It is still using a "majority" measure, but of votes now qualified as remaining after eliminations.
As I've pointed out, if we think that IRV finds a "majority" in all cases, we could just take the election one more round (i.e., eliminate the remaining candidate and transfer votes, if any, for the winner. Indeed, this would be a useful analysis, and it might find a majority, if anyone cared. I'd expect, in San Francisco, that this would have shown a few more votes for the winner. In one election, it might have tipped the balance. So if one actually wants to find a majority with IRV, elimination should be continued until only one candidate remains, not two. However, the point of bringing this up is that, as this point, the candidate remaining has 100% of the last round votes. That's unanimity, if we only talk about uneliminated candidates.
Now, language being what it is, once some special usage is established, ordinarily a word might then be used without qualification. However, here, because of all the issues, it is dangerous (i.e., risks misunderstanding) if the word "majority" is not used precisely, which requires either some special terminology, or that the basis for the majority always be stated. This problem does not disappear by including a special section on it; I suspect that Tom thought it would, and then "majority" could cheerfully be used even where statements would become incorrect if "majority of the votes" is glossed. The reason is that "majority of the votes" is common usage. In a place using runoff elections, that is the language used; no action is complete without a majority of the votes. This is, indeed, very common practice in NGOs, and it's a basic protection against the vagaries of elections with 3 or more candidates. Robert's Rules discourages eliminating *any* candidates, so if no majority *of the votes* is obtained, an entirely new election is held, no candidates are eliminated (but some may and often do drop out, which is one reason why one-vote methods work).
Preferential ballots make it easier to find a majority approving of the result without requiring tactical voting, even with a fair number of candidates, but even with as few as three candidates, as we've seen, it does not necessarily happen. From my perusal of the Cary election, it appears entirely unclear who would have won a runoff election, Maxwell or Franz. It might be predicted from Bucklin analysis, though I'm not sure the data is available. (It was possible in San Francisco because, in 2004, the number of votes in each rank, total, was reported -- this is from optical scan, I think.) But it was complicated by the advice the losing candidate (eliminated before the final round) gave his supporters: write me in at each rank.)
Now, as to what is said about each method, it's POV. With reference to Tom's first draft, above.
Plurality. doesn't require a majority. (But think of it as being rounds. Eliminate the candidate with the lowest vote at each "round," and keep this up until there are two left. The winner is the one with a majority of the votes *remaining.* Of course, we don't think of plurality that way; instead we do batch elimination of all but those at the top, which is only one if there are no ties.)
In two round system, a majority winner is a candidate who has more votes than all others combined in the first round, or more votes than a runner up in the second round. If fewer voters participate in the runoff round, the winner might have less votes in this round than the first round. This is POV, it's an argument comparing two different elections. Two round system requires a majority of votes cast in the first round *or the first round fails*. Then a runoff is held; this is a new election with two candidates eligible to be on the ballot. In fact, write-ins are typically allowed. Because, however, it is considered that at this point enough is enough, I think that a majority is not required. The argument about number of votes in the last round is POV. Sure it's possibly true, but the reverse also can be true. The sock MilesAgain tried to rake me over the coals on that claim, but I'd been looking at Cary, and the runoffs that had been held there. Turnout was typically about the same for the primary and the runoff, and in one runoff the turnout was higher. Consider this: an election where, unexpectedly, a dark horse candidate finishes second, and there is majority failure, so it goes to runoff. What will happen. Suddenly, a whole chunk of the population might see hope where they previously had none, so they are highly motivated to turn out. The dark horse could win. In the San Francisco elections prior to IRV, in the year studied (2000, which was remarkable for a high number of runoffs required -- one of the little hidden pieces of POV imbalance in the article, a very unusual year was reported as if it were typical, though this certainly may have influenced San Francisco voters in 2002), one-third of the runoffs *reversed* the plurality win in the first round. With IRV, no reversals have been seen. In Australia, reversals occur a reasonable percentage of the time. But Australia requires full ranking, plus those are partisan elections. Nonpartisan elections are, I suspect, more likely to see reversals with runoffs, and less likely with IRV, and the effect of voter motivation may be behind it. Eliminating runoffs by using IRV may "improve voter participation," but not the making of decisions by voters sufficiently motivated to vote. In any case, a comment about the justice or injustice of two round systems is inappropriate here in the article. What's a majority is the question this section addresses, and it is not any different than with "plurality." It's a majority of valid votes cast.
In Instant runoff voting a majority winner is a candidate who has more votes in the final round than all other candidates combined. This winner may have a majority of the active votes, but less than a majority of the total ballots because some ballots offered no preferences among the candidates in the final round. Of course, treat this similarly to the treatment just given to two-round system, and we would add, "In IRV systems with limited ranks, which is every one in use in the United States [is this true?], there may be insufficient ranks for voters to be able to vote sincerely and also vote for those candidates." None of this makes the last round plurality a "majority of votes cast." That is, calling this candidate a "majority winner" is misleading.
In a Condorcet method, a winner assumed to be the candidate who can defeat all others head-to-head. Some call this a true majority. This candidate may NOT win an Instant runoff if he is eliminated before the final round. Even this true majority exists, the majority status can be in doubt since a candidate might win all head-to-head comparisons to competitors but fail to win a majority of ballots in all comparisons due to exhausted ballots, just like IRV. Nobody calls the Condorcet winner a "true majority." Besides the bad syntax. That winner is called "the Condorcet winner," if one exists. Condorcet methods satisfy the Majority Criterion (all versions), but truncation ("exhausted ballot" is not relevant to Condorcet methods) does mean that the Condorcet winner might not gain a majority. But it's unlikely. Condorcet methods consider all the votes simultaneously, so it is more capable of finding a majority winner.
In an Approval election, voters are allowed to mark more than one choice and have multiple votes. This means that it is possible for more than one candidate to have votes from a majority of voters, while in single-vote systems, a majority support can only be held by one candidate. That is correct, but, again, not relevant. If Approval is implemented in a jurisdiction with a majority requirement, there may be more than one candidate qualified to be elected by gaining a majority, but the one elected is the one with the most votes. And this candidate, in this event, has gained a majority. That another candidate, or even more than one other, may have gained a majority is moot. It simply means that more than one candidate could have been elected by a majority. I have proposed, actually, that if this double majority occurs, a runoff be held, because it is not clear that a majority of voters preferred one to the other. However, majority requirements exist for Ballot Questions, and if there are multiple conflicting Ballot Questions gaining a majority, the one with the most Yes votes prevails. Approval Voting, actually, with a majority of votes cast required. In this case, the majority of votes cast is the votes cast for each question, not a majority of votes cast for *any* question. (This would apply to Approval elections if they were Yes/No elections, with a majority Yes vs No required for each candidate to be eligible to win.)
No election system can guarantee a majority of voters will support a single winner and the only way to make a promise on majority is to redefine it in a way that exclude voters from the count who refuse to offer a vote among the final set of choices. Well, they do it in Australia. However, other than coercing voters to fully rank all candidates, which is what they do (by considering ballots spoiled if not fully ranked), it's true: there is no way to guarantee a majority in any election where more than two candidates are allowed. Top-two runoff *almost* guarantees a majority, but typically top-two runoff does allow write-in votes, so, theoretically, it is possible for a candidate to have a plurality and not gain a majority, due to write-in votes. My guess is -- I have not checked -- that majority is not required in the second round of top-two.
If an election implementation requires a majority winner, then it must be defined in a way that limits a winner from being declared and leaves the election result open to a further election process. This is a bit confused. It'd put it: "Generally, if law requires a majority winner for an election, the election may fail to find such a winner, requiring further process. The only know ways to avoid this are to limit the candidate set to two, or, with preferential voting and more candidates than two, to require all voters to fully rank all of them."
Some elections, including variations in IRV implementations, may allow voters a chance to offer a protest choice like None of the above that rejects all choices offered which may binding or nonbinding. If nonbinding it has no effect on the election. If it is binding this vote can block a winner from being chosen. While this is certainly interesting to consider, I've never seen such an IRV variation. Has anyone? NOTA, if nonbinding, means nothing, what's the difference between this and a blank vote? If the vote is counted, is it part of the basis for a majority? If so, then it is binding if a majority is required. Now, in most elections here in the U.S., a write-in vote is allowed. One could write in any name, and, as long as this is the name of an eligible candidate (which usually means a resident of the jurisdiction, of the age of majority -- a different meaning entirely! -- and not disqualified for some reason such as citizenship or felony conviction), it serves the NOTA function; if enough voters do this, it would deny a majority to the winner. But this is all theory. Majority NOTA *should* cause election failure in a democracy. But the meaning for an officer election is unclear.
And ... a huge pile of original research, unsourced. Some of it could be sourced, to be sure. Some, quite possibly not.
By the way, I appreciate the tremendous effort Tom has put into trying to clean up this article. Regardless of whether or not I agree with all the changes he has made -- and it will take some time to catch up to his racing ahead -- his work is appreciated. -- Abd ( talk) 06:30, 4 February 2008 (UTC)
(unindent) Thanks, Mr. Bouricius. It's a little offensive, I'll note, to project upon me POV "fixation." But, hey, I've seen worse said around here. I do not have any intention for majority to mean a number as large as possible; rather, I'm focused on it meaning exactly what it has always meant when we say what is in legislation and constitutions as majority of the votes. I used the wrong number in my original presentation of the data from San Francisco. At one point I also incorrectly asserted that blank ballots might be part of a majority basis, but both these were simply errors. I make mistakes. They do not proceed from any intention to deceive or exaggerate. (As to the ballot issue, "majority of the ballots returned" would be the term, and "majority of votes cast" would be something different, quite clearly, since blank ballots don't contain a vote. Further, there is a legal presumption that an invalid vote, not for an eligible candidate, isn't a "vote." It is the same as a blank.
Thanks for the information about Burlington and Takoma Park. Full ranking is, of course, a fish bicycle in Takoma Park, but I think there were more than three candidates in Burlington. IRV may indeed be lumped together by some as a "majoritarian" method, but this may merely reflect two things: first of all, it satisfies all the definitions and proposed definitions of the Majority criterion, and it does seek a majority. However, it does not maximize majority, for the sequential elimination, for reasons Mr. Bouricius surely knows. That is, it is possible that one may derive, from the votes on an IRV ballot, a winner who would be preferred by an actual majority over the IRV winner, and over every other candidate, and who would, indeed, have a majority of ballots containing votes. For example, suppose there is a set of voters and preferences:
Far from being a weird contrived example (though the extremity of it, the almost perfect balance, is indeed chosen for purpose of illustration), this example represents a polarized electorate, with society divided almost exactly in two (51:49) over some issue or constellation of issues. A represents, perhaps, the middle of the left, C represents the middle of the right (perhaps they were chosen by the Left and Right parties, and B is a compromise candidate, as Robert's Rules surely considers important. B is in the center. So half of B's supporters come from the Left and half come from the Right. If B is eliminated, A wins, narrowly. But 65% of the voters preferred B to A. This is the classic Center Squeeze. B is the Condorcet winner, but because the B "party" -- if there is a party -- is slightly less popular than the two extreme parties, B loses. It is entirely due to the elimination process. Top-two runoff, of course, has the same problem. This is why Robert's Rules dislikes candidate elimination.
Now, facing this, if some of the C voters realize their predicament, they *might* vote, instead, B>C>A, thus turning the election to B, whom they *greatly* prefer to A. However, Favorite betrayal criterion is not the topic here. Consider how Bucklin handles this election: in my opinion, in Bucklin, the vast majority of voters will vote sincerely; and whether or not truncation will be practiced quite depends, I expect, on preference strength. However, if we assume that some of the A voters and more of the C voters realize the risk of truncating, plus the middle third of the B voters truncate (they don't really care that much who wins, A or C, if B doesn't), We might get
(Much of the information and analysis we have comes from FairVote and, in my opinion, it has not been objectively reported, original sources aren't available for checking, etc. The claim that Bucklin voters did not add additional ranks because they didn't want to hurt their favorite is unsupported, even though that phenomenon can be shown to be, in pure theory, strategically advantageous. In fact, the dilemma only arises in partisan elections where a voter is puzzling over whether to vote for Bush and Gore, or just one of them. For third party supporters, where their favorite has little or no chance of winning, it is entirely moot; these are the voters who will add ranks, almost certainly, and some major party supporters will also, where they want to send a message to their own party as to which way to lean, and would not be seriously displeased if it *did* turn out that their vote caused the third party candidate to win. But, by definition, *most* voters do not have a strong motive to add ranks, and that is harmless. But, of course, it *could* cause majority failure. But very unlikely to do so. In nonpartisan elections with many candidates, majority failure becomes very possible, even likely, unless there is a runoff reducing the candidate set (or even not: voters change their minds, decide to compromise, and candidates drop out). But it's worth noting that Bucklin analysis showed a far higher expression of a majority result -- though certainly the argument that voters may vote differently with Bucklin should be noted, though I suspect that, in fact, they won't. Not in large numbers, in nonpartisan elections, except for the minority of voters who are truly dedicated to one candidate. Lots of voters, in the Smallwood elections, did add additional votes, and they turned the result, hence the suit by a supporter of a loser, who had enjoyed a plurality in the first round.)
Now, the situation above produces the same results in IRV as the original fully-ranked results, but, because of the truncations, without a majority, showing something important. The truncations that matter are the B voter truncations, and quite a few of these, actually, can be expected. If we assume a linear issue space, with B in the middle of A and C, voters near B have little predicted preference between A and C, and so may very well not vote for either one; this, indeed, would result from, in Bucklin, a fear that their vote for A or C might cause B to lose, and that is valid, and if they act on it, it may improve the overall result' (from a voter satisfaction standpoint, assuming that relative satisfaction for each voter is as important as that for every voter).
But in Bucklin, because no votes are discarded (unless they are not reached because a winner has been found before that rank is added in), we have no majority in the first round, but A has a plurality. In the second, we end up with A:46, B:53, C:44. B wins with a majority even stronger than the A majority in the original IRV election: The vote counts add up to more than 100, *but* B has a vote from 53% of the ballots, which is entirely accurate. B is a majority winner, and Bucklin has been more efficient at finding that, even in the presence of truncation. What if there is no truncation (which is highly unlikely)?
That's easy. The results are, from the original IRV election, using Bucklin analysis: A:51, B:100, C:49. Truncation does conceal the true support for B, which is actually spectacular. Bucklin can fail to elect the Condorcet winner, to be sure, but it looks like it does so more often than IRV. (There is data on this from Warren Smith and, I think, others, based on simulations). Bucklin finds *every* vote, it does not discard any of them. If voters don't truncate, every voter would vote for B as being in the top two; there might be more candidates in the race, but they were left out for simplicity, these were the rankings of the top three.
So, sure, IRV is considered a "majoritarian" system, but what, exactly, does that mean? What is the definition, what criterion is applied to determine membership in the category? I did not notice either Richie or Bouricius telling us; rather, all I've seen is argument deflecting the point using the fact that "majorit" is found in both words.... and, of course, both are related to "majority" in some unspecified way. IRV does find a *kind* of majority, and some forms of IRV always find that kind of majority: full ranking does it. But IRV does *not* find -- and does not do a good job of seeking, because of vote dropping -- a "majority" as a "majority of votes cast" as it has always been defined, and as it is defined in Vermont. In future campaigns to implement IRV, I predict, this fact will not escape notice if the jurisdiction is using top-two runoff, because those jurisdictions would probably not be using top-two runoff if they did not value verifying a majority of votes cast, and I don't think arguments based on it will be entirely missing from the voter information generally sent to voters, nor will the proposition summary be misleading as it was in San Francisco. So San Francisco might not be repeated, and we may be seeing the peak of IRV adoptions. It is also going to be, I predict, more difficult to find the funding for implementation, if inconvenient questions start to be asked about these issues. So there is my political punditry, such as it is. I've often been wrong. All I know is that the facts uncovered here have not escaped the notice of people working in the field, there are very active opponents of IRV out there (opposed to it for a host of reasons, sometimes they are strange bedfellows), and I don't think they will be caught sleeping again.
Meanwhile, sourced facts, relevant, are not to be removed from the article by editors with a COI, or even mere POV editors, based on vague claims that it is "too much detail," or spurious or off-the-point claims that it doesn't affect "real elections." IRV is not only about actual elections, it is a method and it is being considered for application in many places, political and nonpolitical. An article about the method should consider relevant aspects of the method, and, last time I noticed, Vermont was a notable state. I suspect that the conflict between IRV and a "majority of votes" is also going to affect many other jurisdictions considering the issue, even if they could constitutionally discard the "majority of votes" standard. The only difference with Vermont is that the IRV promoters -- Mr. Bouricius introduced a version of the legislation (the first?) -- had to respect the difference, they could not pretend that the last round was a "majority of the votes," they knew, presumably, that this did not have a snowball's chance of making it past the courts.
When I originally raised this issue here, Mr. Bouricius ridiculed the distinction I was making based on my understanding of Robert's Rules and thus my interpretation of the passage there describing a form of "preferential voting" that is very similar to IRV -- or even identical. Yet, in fact, he knew very well that there was a basis for my objection. I do, by the way, trust his good faith, I don't think he was deliberatively concealing what he knew. I think that, rather, he is as he was, a politician, who thinks in certain ways, and that kind of thinking is why we restrict what Conflict of Interest editors can do, because that kind of thinking is extremely common with them. It's also the norm, in fact, among POV editors of any kind, and my pots are black just like almost everyone else's. I do, however, understand the difference, at least in theory, between my opinions and the truth and what is NPOV. (Those are three different things, but NPOV is as close as humans, absent direct mystical experience -- debatably --, that humans get to expressing the truth.) To find NPOV, I usually need collaboration with others of differing points of view. And to terminate this fugue, some seriously contentious people have learned to work together by understanding that "For our group purpose, there is but one authority, [the truth as it may express itself] through our group conscience." That's a 1940s term for "consensus."
My goal for this article is that all reasonable people with whatever point of view will say, "Yes, that's accurate" and that they will consider it interesting and informative as well. -- Abd ( talk) 20:57, 4 February 2008 (UTC)
(unindent) Thanks, Tom. I do think there is a way to make the article more clear. A section on the issue of majority and IRV may well be a solution; there are some problems with sourcing and original research, but it's possible those could be overcome. But simply settling on some standard terms, defining those (with source), and then using them uniformly where possible, could be sufficient. The point about all this is that the issue of majority is, in fact, an important issue where IRV is being considered. The major elections in the U.S. have been ones where IRV has been implemented as a replacement for top-two runoff, and there has been a great deal of misunderstanding -- and misrepresentation -- about this. So we should simply be clear: what does IRV do and what does it not do? How is it performing? What can be reported from reliable source? There is also the question of where to put some of this. The San Francisco voter instruction pamphlet language from 2002 is important, probably it should be in the Controversy article. The results in terms of whether or not a majority of the votes was found for the winner are relevant, and those facts are available in official reports. We could more efficiently report a summary of all this, but we will, at this point, have trouble finding one. I intend to fix that, actually, but that will take some time. (It involves peer-reviewed publication.) (And, of course, if I'm the author of a paper, ... up creeps Mr. Conflict-of-Interest, perhaps.)
As to executive summary, well, some good writing has been done when someone asked me for such. I tend to respond to requests, it's part of the disorder I mentioned above. I think that if my hangman forgot how to tie the knot, I'd show him
-- Abd ( talk) 03:30, 5 February 2008 (UTC)
Someone inserted information about a survey of disabled voters as it relates to the change in Scotland to STV. This is not appropriate for this article, fundamentally because Scotland did NOT switch to IRV. The editor stated falsely that STV is the Scottish name for IRV. There were numerous changes enacted...such as from single-member constituencies to multi-member (electing three or four seats per district). The fact that STV uses a ranked ballot is only one of numerous changes, and it is wrong to claim the survey was about about IRV. Tbouricius ( talk) 22:50, 4 February 2008 (UTC)
Yes, I'd agree with Tom's suggestion. At the same time, I must point out that User:Tbouricius has, here, made a contentious edit. He should know that, as a conflict of interest editor, that's not allowed. It's absolutely fine for him to argue his points here, but not to revert other editors. His argument has some merit, but it is by no means conclusive. The actual document referenced specifically mentions the ballot form as being a problem, thus the objection could indeed apply to IRV. I edited the reference to make it clear that this was an STV election, and I thought I put in that it "used the same ballot." And the procedure is the same. These were ballots requiring voters write in numbers. That could be difficult for some disabled voters, more difficult than marking a box, and it creates quite a few opportunities for error, such as writing the same number twice. I'll also note that Bouricius totally ignored Ask10questions' quite on-point evidence that IRV and STV are being considered the same; indeed, didn't we just go through a whole controversy over the introduction, with the sock puppet MilesAgain and Bouricius insisting that IRV uses a "single transferable vote," and, indeed, we all know that IRV is simply single-winner STV. Same ballot, same considerations for the voter, same counting method for finding the first winner. STV only deviates in how it proceeds when then looking for the additional winners. All a voter needs to do, with both methods, normally, is to vote for the favorite candidates, in order of preference. However, the voter now faces some consideration of how many candidates to rank, and what to do with all those candidates. With FPTP, the voter simply ignores unknown candidates. I'll agree, that's fine to do with STV as well, but ... does the voter know this? The example in the voter information image I cited shows full ranking. I can easily see some people getting confused. Is this a reason not to do IRV? Probably not the strongest, but it does point out the need for education, which is an additional expense involved (San Francisco spent quite a bit, it's part of the conversion cost.) -- Abd ( talk) 04:28, 7 February 2008 (UTC)
On the other hand, I thank Mr. Bouricius for inserting the reference for the claim regarding Bucklin strategy. I did not find text for that reference, not yet, and it's getting late, but I did find another historical article on the system being used there.
[11]. And there is quite a bit of talk in this about voter strategy. It's a complex topic, and one that I don't think belongs here, unless it is going to be explored neutrally, which would require explaining *other* ways that Bucklin and IRV differ. The argument about Bucklin encouraging voters to bullet vote is, in my view, like many FairVote arguments, not wrong, but misleading. Voters who have a clear favorite simply vote for their favorite, unless they think that favorite isn't likely to win, in which case they add additional votes. The so-called Burr Dilemma is a dilemma allegedly faced by candidates, in how they would recommend voters vote. It doesn't afflict voters nearly so much and, I suspect, most voters will cheerfully ignore it. And the Burr Dilemma is named after a situation that has been called "Approval voting," but that was simply the invention of the author of that paper. It was not Approval voting.
Alabama, it appears, allowed two votes, a first rank and a second rank, and it's legitimate to call it Bucklin because the second rank votes are added in. And apparently, voters did add them in. From the source I've given:
That's fairly silly, actually. The Bucklin results would separately show the first and second rank votes. The sum of all votes is not a meaningful measure in Bucklin, rather the vote percentage, showing true support, would be in excess of 80 percent, almost certainly. That is, this author misreported the results. (But because they did not demand a majority, the Alabamans may have routinely done the same thing: if no majority in the first round, the second round votes are added in and the winner is the one with the most total votes. "Most total votes" will be the same a "highest percentage," no matter what the basis for calculating it is. But if we want to understand what percentage of voters voted for Black, we won't consider the total vote as the basis, rather we may assume that the sum of all first rank votes shows the number of ballots (I'll neglect the possibility of blank or invalid first rank and then valid second rank.)
Since Black got 80% of first rank votes, he must have received at least that much support from the voters. The *minimum* would be 80%; he would get that if none of the voters who voted the other 20% of ballots voted for him in second rank. Let's suppose that the opponents' supporters all bullet voted. In order to reduce his percentage to 68%, the basis must have increased to 80/68 or 117.6%, so 17/80, or 21%, if Black voters, with Black being a very popular candidate, obviously, nevertheless voted second rank for another. The article elsewhere goes on to note: "Second-choice voting also made it possible for a candidate to win with fewer first-choice votes than an opponent." It considers this a defect, but why it's a defect is unclear. IRV can do the same. But it apparently doesn't do it very often; on the other hand, Brown v. Smallwood was a case where it happened, and Brown was a supporter of the opponent of Smallwood who had been the plurality winner in the first rank votes. Apparently the Minnesota Supreme Court also considered this a defect, and it is the basis, essentially, on which they tossed out Bucklin. I.e., exactly the same thing that IRV does. The argument about second rank votes hurting first rank choices is possibly mentioned in Bucklin, but the bulk of the argument from the court is over this impairment from other voters, that they might turn a first rank plurality into a loss. Which, of course, was the whole point of having a reformed voting system. If it never did, why bother?
Now, if we are going to have what is presented here as an argument against Bucklin, in this article, then we will need to have balance. Is that what these editors want? Because it can certainly be done -- but when I've put such in the article in the past, it was vigorously resisted. Until we have some consensus on this, I'm taking the reference to strategy out, it is a complex and controversial issue; but I'm certainly very interested in seeing that source, and in the history of Bucklin in general. It's the closest system we've had to Approval in public elections in the U.S., as far as I know, at least in the Duluth implementation (which really did allow voting for as many candidates as you wish -- in third rank). It's really a ranked Approval, giving first shot to first rank choices, then expanding until a majority is found -- or it terminates with only a plurality. From what I've seen, it would probably be more efficient than IRV at finding a majority unless voters massively bullet vote, more than they are currently truncating in IRV. And this is a huge and distracting can of worms here, I'm trying to stuff them back in at the moment! -- Abd ( talk) 05:13, 7 February 2008 (UTC)
![]() | This page 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. |
Archives: ~ Jan 24, 2008 - Feb 13, 2008
An anonymous editor took out the entire San Francisco text, with a cryptic edit summary, which I'm sure will make sense after it is explained. There is a lot of data in that text, the editor did not explain what, specifically was wrong. But there is only one percentage reported as such, so if that alone was wrong, it should have been corrected, not removed, or, at least, not the entire report. I've reverted, and invite this editor or any other to find and explain errors. These facts were reported here before being put in the article and are fully supported -- except for any errors, of course -- by the source. So, please, if there are errors, *fix* them! It is not my personal responsibility, that is a misunderstanding of what Wikipedia is. It is not a debate forum. -- Abd ( talk) 15:33, 24 January 2008 (UTC)
header 1 | header 2 |
---|---|
data a1 | a2 data |
data b1 | xx data b2 |
Now that the U.S. impelmentation mess has been moved to a new article, I am not sure if this is the best page to explain some errors in Abd's analysis, but the new article doesn't seem to have a discussion page yet. So... One place where Abd has simply gotten it wrong with his "original research" is his reporting on San Francisco. He has misunderstood the way San Francisco reports the results. What the city does is treat people who skip over the race entirely as "undervotes". That means someone could win 55% of votes in the first round, but Abd would still call them someone elected with a minority of the votes. It is the standard, normal, and correct that percentages are always reported as a percentage of votes cast IN THAT RACE, rather than as a percentage of votes cast in other races (such as for president). The drop-off for lower offices is typical. The fact that most jurisdicitions use a combination ballot paper with multiple offices does not alter that correct way of reporting percentages.
Take the 2005 citywide race for city-assessor. Phil Ting won 47% of the first choice vote. He then won substantial number of second choices from another Asian candidate and won with an absolute majority of the first-round vote. But Abd has him as not winning a majority of the vote, because he counts in people who didn't vote in the race at all. See:
http://www.sfgov.org/site/elections_index.asp?id=61607
Similiarly, in 2006, Chris Daly won in the 5th round of counting with a majority of continuing votes. They stopped the count at that point with him something like 4 votes away from a majority of the valid first-count vote, so one could argue he technically didn't win a majority of all votes cast in the race, but they stopped the count with his three strongest opponents still in the race, because he had won a majority of continuing ballots (discounting exhausted) -- meaning he obviously would have won a majority of the first-round vote if they had reduced the field to two.
It is true that the hotly contested race in District 4 was won by a majority in the last round, but that majority was less than 50% of the valid first choice vote. That also has happened in a couple other Supervisor races, but people have accepted the results just as much as they would have if someone ran a runoff with the voter turnout a little lower in the runoff. It is never proper to describe the percentage in the runoff by using the denominator from the first round of voting.
Abd really messes up the race he profiles in the write-up - -the one with 22 candidates. He suggests the winner had only 37% of votes, but he skips over the MAJOR fact that the count ended with three candidates still in the race -- the winner had a majority of the final round vote with two other candidates still in the field so they didn't need to eliminate the third-place finisher, because no other candidate had any mathematical chance to win. So Abd is really comparing apples and oranges. True, Mirkarimi would not have won 50% of the first-choice vote if they had gone to the final two, but he almost certainly would have been a good bit over 40% -- and his point just is highly misleading and does not belong in a NPOV article. Tbouricius ( talk) 21:20, 26 January 2008 (UTC)
Abd. You might want to redefine the academic consensus that runoffs and instant runoff voting are majority voting systems, but you know you are doing so outside that academic consensus. Runoff elections also can have significant changes in turnout -- when one of two rounds can be decisive, significant differences in turnout obviously can raise questions about whether it was a "majority result", but the academic consensus is that it is. Choosing to make your case within the Wikipedia article that it's not is a highly political decisions, reflecting your own point of view.
In San Francisco, ithin the city, there was absolutely no controversy about the result of the Mirkarimi win in 2004. People were very used to candidates winning majorities in runoffs with a drop in turnout, and that's how this was seen --- with the added fact that the turnout in a real December runoff would have been much, much larger. Mirkarimi had a maority of continuing votes, people knew those were the rules, and he was generally seen as winning a majority in the instant runoff because he was the majority choice of the top candidates. You obviously can dispute that based on your own numbers, but it's one definition of majority versus another.
Furthermore, your selective highlighting of the 37% number in a short summary of data is highly misleading when there were three candidates left. Did you not know that when you did this or chose to overlook it?
As a final point, every winner in the thousands of election in Australia's house of representatives has won with an absolute majority of the valid first-count ballot. That's a product of their rules, which invalidate ballots that don't rank everyone. So if one wants absolute majorities of first-round ballots, it can be done through such a requirement. But if voters don't want to rank everyone, allowing them to truncate their ballot is reasonable policy decision. It also has been seen as a reasonable decision to give voters three times the chance to cast an effective vote with IRV as they would have with a plurality ballot, although most IRV backers would prefer a ballot with unlimited rankings (as indeed will be what San Franciso uses when it has voting equipment that allows it). —Preceding unsigned comment added by RRichie ( talk • contribs) 14:35, 27 January 2008 (UTC)
User:Tomruen moved the detailed election information to a new page he created. There are now *three* pages on the history of IRV, and none of them really address much of the history! In the process, without comment, he removed the election analysis that summarized results from IRV elections so far.
I think there is a basis to have subarticles. One, a history of preferential voting (using the STV model). This is general, not confined to the United States. It would have a section on the U.S., a brief summary. Second, a history for the U.S., so the two pages that now exist should be merged. The other one on the U.S. was created as another POV fork; the creator, a probable sock, simply put back in what had been taken out of this article based on improper sourcing or for other reasons.
The preferential voting/STV article currently is called History and use of instant-runoff voting. This should use a name by which the method has been known for almost its entire history, all but the last decade, rather than the newly-coined instant runoff.
The other U.S. history article is History and use of instant-runoff voting in the United States. And now, of course, Ruen's IRV implementations in United States.
However, each article should stand alone. If there is a history in the U.S. article, it should be summarized here, using WP:SS, and the summary should be NPOV. Among other things, that means that there should be balance. A list of adoptions, which is what Ruen left here, is effectively POV. It's promotion. Most of those adoptions are moot, some may never see an actual ballot.
The forks or subarticles should have names that have Instant-runoff voting at the beginning, to follow the convention that puts the articles together in alphabetical indexes. That's what we did with Instant-runoff voting controversies.
We started out with a list of adoptions in the U.S. I added detail, such as describing which ones had actually been implemented or which were actually scheduled for elections, plus a brief summary, when the information was available, of how IRV was performing in terms of its goals. IRV is being implemented in jurisdictions where, previously, there were top-two runoffs. So how it performs in comparison to top-two runoff is highly significant. But, I can imagine, it was not liked that this information was being noticed, so ... with Ruen's move, it was taken out. Nice work, Tom. -- Abd ( talk) 05:01, 27 January 2008 (UTC)
Abd - you added selective detail. If you're going to write about the specifics of IRV elections, it makes sense to be much more comprehensive than these articles have been. Putting it in a separate article that is flagged as incomplete makes sense, doesn't it, rather than cherrypick examples? —Preceding
unsigned comment added by
RRichie (
talk •
contribs) 14:37, 27 January 2008 (UTC)
Still learning Wikipedia here - I'll start signing these.
Also, Abd, I had a concern about what you're doing with these edits on San Francisco, etc. Aren't you doing "original research"? Wikipedia says that A"aticle policies" include "no original research", but that's exactly what you're doing. Your mistakes are exactly why I assume original research is frowned on. I think you should put this kind of information on your own website (Center for Range Voting, which I know you post things too). RRichie ( talk) 16:57, 27 January 2008 (UTC)
I moved two talk sections to Talk:IRV implementations in United States HOPING this will solve the unnecessary expansion of details in the main article. I also expanded the bullet-list format there to sections so each city can have extended information as details and results progress. Please move all discussion there. Tom Ruen ( talk) 07:10, 28 January 2008 (UTC)
I just removed the {{ POV}} tag from Approval voting, because all of the concerns in the section discussing it have apparently been addressed.
For here, I'm not so sure. I know that we have had some serious disagreements, but things look to have settled over the past week pretty well, at least as far as the neutrality concerns that I recall are involved. It's hard to tell because all of the POV tag discussions seem to be archived.
Are there remaining neutrality issues, and if so, what are they all? MilesAgain ( talk) 02:16, 28 January 2008 (UTC)
Section talk moved to: Talk:IRV implementations in United States Tom Ruen ( talk) 07:03, 28 January 2008 (UTC)
I removed this paragraph from implications section:
I didn't like the soft some other voting methods (What methods!? Bucklin or Borda?). I didn't like the soft could prolong the counting procedure (compared to what?) The entire paragraph looks to me like whining. I'm content if something is restored that is more objective, perhaps:
BUT basically the two examples in the final paragraph already said it better!
I know IRV-opponents want to use this summability limitation as a reason to discount IRV, and happily claim whatever alternative rank method Borda count or Bucklin voting or Condorcet method or whatever doesn't have this limitation. BUT you can't just say A is bad because there's a small cost, while the cost to alternatives is neglected (i.e. Voters don't WANT lower choices measured AT ALL unless their choice is eliminated, SO it's the nature of runoffs themselves.)
I can't really support ANY comparisons to other methods under implications, since implications ought to refer to changes from what IS, not alternatives which do something completely different.
Anyway, I figured this was as good of a fight as I'm willing to offer for now. Tom Ruen ( talk) 02:36, 30 January 2008 (UTC)
P.S. The Summability criterion was one of the articles deleted from Wikipedia, and reconstructed on Abd's user subpage: User:Abd/Summability criterion.
P.S. If Electoral fraud is an issue with sources, perhaps it ought to be included. I accept compared to plurality, it's a legitimate issue to consider. I mean it seems much bigger - like all votes should have a paper trail, and ballots should be designed in a way that allows fraud to be detected. For example, a checksum could to be performed at each precinct, counting how many marks there are at each ranking (like Bucklin sort of), so if marks are added (or removed?) after the voter is finished, it can be detected. I don't know ANY sources for such concerns, but it seems obvious to me that people in-the-know must think about these issues. Tom Ruen ( talk) 03:00, 30 January 2008 (UTC)
User:Peter Ballard added the following text (in my italics) to the article:
Now, the problem is that this text draws a conclusion. First of all, the facts are going to need a source. However, the "facts" don't make the conclusion. Somebody must. Who? It's fairly clear that the National Party exists, and has for a long time. However, if I'm correct, it's dying out. It used to be a more or less equal partner with the Liberal Party. If it is dying, this is *confirming* similarity with plurality, rather than being a "significant difference." IRV, pretty clearly, allows far more flexibility in voting, *theoretically* allowing third parties to function without spoiling elections due to vote splitting, but it should be noted, if third parties run on a par with the two major parties, a nastier effect can appear than the simple spoiler effect, there is "center squeeze," a phenomenon which is rare when there are only two dominant parties, and I think this is behind Duverger's law. (When Center Squeeze is taking place, IRV is discarding the candidate actually favored by a majority over either of the other two major candidates, but with fewer first preferences, which is essentially behaving like plurality.) There are other methods of Preferential voting that might not have this problem. The *real* problem is, of course single-winner, which is, in my view, very inappropriate for representation, even if the method works and the majority votes for the winner, it leaves a minority, sometimes almost half the electorate, unrepresented. That's seriously defective. "No taxation without representation" was a founding slogan for the United States. Okay, who represents me? Am I represented because someone else appoints my representative? If so, what were the American colonists complaining about? They had representatives appointed by the King!
(To be sure, "representation" was thought of in those days as being representation of communities, not of people. But, again, it was the people who were taxed, not, generally, communities.)
If the survival of the National Party is to be mentioned, *how* it is surviving, in what numbers, with trend over time, would need to be mentioned as well. People, in the U.S., are interested in these things, sometimes, because they are supporters of third parties, and would like to know what the ultimate effect is. If the ultimate effect is that the party dies a slow death, that's quite different than "thriving," and isn't even fully "surviving." Dying takes time. Is there any example of a third party *arising* under IRV, and growing to become a major party? *That* would be interesting, absolutely. -- Abd ( talk) 17:25, 30 January 2008 (UTC)
Is it really controversial to suggest existence is aided by a lack of need for vote-splitting from a third party supporters and a similar major party competitor? Tom Ruen ( talk) 21:42, 30 January 2008 (UTC)
-- Abd ( talk) 22:02, 30 January 2008 (UTC)
Removed:
This statement is without proper context since the only comparisons that will effect parties and candidates are plurality or two round system which have the same behavior! Tom Ruen ( talk) 22:36, 30 January 2008 (UTC)
IRV elects candidates who have a majority of votes among those voters who express an opinion between the two finalists. Abd repeatedly seeks to insert the point that IRV winners may have less than 50% of ballots that had a vote for a candidate in the first round. Whether such candidates are "majority winners" is important to his POV. This distinction appears at various points in the article, and probably should be reduced to one statement early in the article so as to avoid having to repeat the concept over and over. For example, Abd has inserted the unique example of a Vermont bill in which exhausted ballots can result in "no election" by the voters. I know about this in detail, because I wrote the bill. This language was crafted to fit with the language in the Vermont constitution, which says there is "no election" if no candidate has "the major part of the votes." If there is "no election" then the constitution charges the legislature with selecting a governor. While that language is not crystal clear, so that exhuasted ballots might still be discounted, there is a principle of constitutional law that legislation that makes a constitutonal provision meaningless is unconstitutional. Therefore, to avoid making the supplementary role of the legislature in selecting a governor a theoretical impossibility, the bill was crafted in this unusual way.
I cleaned up Abd's sentence to give it a citation and explain just a bit...However, I think the sentence Abd seeks to include citing the rare case where exhausted ballots can result in "no election" is so unique, and only exists as a bill and has never been implemented anywhere in the world, that we know of, means it should not appear in the Wikipedia article since it distorts the general understanding. Government elections are assumed to be decisive, with the option of None Of The Above (or exhausted ballots) leaving an office vacant, or resulting in a new election, existing only in theory. I have not deleted Abd's insertion about this unique proposal in Vermont yet, to give others a chance to look it over first. Tbouricius ( talk) 15:41, 31 January 2008 (UTC)
User:Tbouricius edited (properly) my mention of the Vermont situation with IRV, with the summary, "(Changed Implementation" to "proposal". and added cite. However, I think this whole sentence should properly be deleted as a rare exception.)"
Here is why it is actually very important. IRV is (almost?) exclusively being implemented in jurisdictions which have majority election requirements, and therefore, sometimes, runoff elections, which are allegedly expensive, providing an economic incentive for implementing preferential voting, such as IRV -- (and only recently has there been much agitation for any other form or runoff-avoiding reform). The most notable, of course, being San Francisco. Jurisdictions have those requirements because, apparently, it's believed that a majority of those voting should consent to an election. This is, as Bouricius knows, a common belief, supported by Robert's Rules. In San Francisco, IRV was sold with an assurance that "a majority would still be required." Yet, in fact, a majority of those voting is *not* required. The ballot summary was misleading. Voters show up, vote for an eligible candidate, and yet their vote is disregarded when it comes time to determine a "majority." That is a blatant violation of the underlying principle.
It is often argued that, well, if voters don't show up and vote in a real runoff, their votes don't count also! And it's implied, as this article has implied from time to time, that exhausted ballots are the voter's fault or choice. Perhaps, sometimes, but perhaps there were a lot of candidates (more than three) and the voters used all the available ranks by voting sincerely. Or some didn't understand the method, or really detested all the remaining candidates. The fact is that, in the large majority of "instant runoffs" we have seen since San Francisco took the plunge, a majority of voters showed up and voted for candidates other than the winner, not for the winner. If a runoff were actually held, we may suspect, roughly one-third of the time the election would reverse from the plurality winner in the first round. (examination of real runoffs, which will show similar voting patterns in the primary, shows this or something similar, plus we have some analysis on this point from Australia). Some of that may be due to differential turnout, but ... quite likely not all of it (and differential turnout doesn't apply to the Australian data, of course).)
(And if that last-round definition was the meaning intended by "majority," it was thoroughly deceptive, for there is no "requirement," IRV defines the winner as having a majority simply by discarding all votes not for the top two. As I've pointed out, we might as well take it one more round and call it "unanimity." It's not a *requirement*, it is a tautology.)
IRV, as implemented in San Francisco, has defeated the purpose of requiring a majority. By passing the proposition, voters, quite likely without realizing it, discarded the requirement, substituting something else that was sold, through the use of the politically designed name, "instant runoff voting," as being equivalent.
Now, I'm sure that Vermont legislators and IRV activists would do the same thing if they could. What is unusual about Vermont is that the majority requirement is in the state constitution. They can't simply legislate it away, sweeping the issue under the carpet, as was done in San Francisco. So, yes, it is rare, but only because a constitutional requirement for a majority is rare. Many jurisdictions have that requirement, so the idea of requiring a majority isn't rare.
Apparently, this argument has mostly escaped the opponents of IRV. Nothing like this was brought up in all the ballot arguments provided in the San Francisco proposition, and, to my knowledge, it hasn't been raised elsewhere in real campaigns yet. I'm sure that is about to change. This argument, and the data behind it, is now becoming known to activists working on the issue, I don't think it will continue to be possible to sweep this under the carpet. An ounce of experience is worth a pound of theory.
In any case, I'm not suggesting that this argument be put into the article. However, the underlying facts exist and can be properly sourced. They were put into the article, and removed by User:Tomruen as part of his creation of the U.S. history fork. I'll be bringing them back when I have time. The summary of how IRV is performing belongs in *this* article, and, if anyone claims it is cherry-picked, then they are welcome to add the allegedly excluded facts. Such as, for example, if anyone wants to do the searching, how IRV has affected turnout. Certainly that would be relevant and interesting. As well, how has IRV affected election costs, overall. Problem is, it's hard to get that data. Still, if someone can find facts relevant to it, they should be in the history article in some detail, and in this article in summary form. -- Abd ( talk) 20:36, 31 January 2008 (UTC)
I added a test table on the largest IRV election from SF, 2004, could use some comments - Is it useful to include in the IRV_implementations_in_United_States article? Tom Ruen ( talk) 02:21, 1 February 2008 (UTC) Talk:IRV_implementations_in_United_States#San_Fran_results
I think so. I'll comment there. Nice work. Tom, you will learn more if you look at the Excel spreadsheets they provide. These give the full ballot data, summed, i.e., we know how many voters voted for each candidate in each rank. If this had been a Bucklin election, more of the candidates would have reached an absolute majority, and, of course, we'd have all the relevant data. As it happens, Bucklin would have given the same results as IRV in all these elections. (Which isn't surprising. The differences arise in unusual cases.) -- Abd ( talk) 04:10, 1 February 2008 (UTC)
User:RRichie removed the following italicized text from the section on practical implications: diff
The statement preceding it is not true under the conditions that a majority is required. Every jurisdiction actually implementing IRV (except I have not checked out Takoma Park, I assume Mr. Richie could tell us about that one, I presume he sold them the fish bicycle) did so to replace majority-required elections, and IRV was promoted as a way to get majorities without holding a presumably expensive or inconvenient by-election. Indeed, it can do this, but it can also fail, and, it turns out, it is *usually* failing. The reason why a majority is not required in the current IRV elections is that IRV *replaced* the majority requirement, it did not satisfy it, but this appears to have escaped the notice of those who might have been opposed. The voter information pamphlet for the San Francisco proposition did not even mention the argument (among all those who paid for opposition arguments); and to the contrary, the summary language, supposedly neutral, explicitly claimed that candidates would "still be required to gain a majority of votes." (I think that's an exact quote, but it's from memory, might have been a little different. I quoted the exact language in Talk; in a section that was moved to the Implementations article). "Majority of votes" has a common understanding and meaning: take all ballots containing a vote for an eligible candidate and count them. Divide that number by two. The winner must gain more votes than that. The "majority" in IRV is a majority of a reduced subset.
I consider it highly likely that, in the future, new IRV initiatives in such jurisdictions, in order to pass, will retain the majority requirement, for the cat is out of the bag. In my view, IRV with a majority requirement is substantially better than IRV without one. Without the requirement, IRV creates a "majority" by discarding all votes from consideration that don't rank the top two; this results, necessarily, in a majority *of votes remaining*. Take any election, with any method I can think of, and discard all votes not for the top two. Presto! A majority for the winner. Indeed, if IRV can be claimed to guarantee a majority winner, we might as well claim that it always finds a consensus winner, elected by unanimous vote. Just continue the elimination process one more step, until there is only one candidate instead of two! No, IRV does help find majority winners, that is true under some conditions. But it can also miss them, because of the quirks of sequential elimination, as I'm sure Mr. Richie knows quite well. Indeed, he sometimes argues that this is a feature, not a bug. IRV requires candidates to have a certain minimum "core support." Bottom line, this means that the winner, generally, must be one of the top two in first preferences. (With batch elimination, used with some elections that Richie cheerfully calls "IRV," that is exactly it). Apparently, this is more important than having winners actually approved by a true majority of those who bothered to show up and cast a valid vote. I.e., standard majority rule.
Since we are on it, there is one standard response to this line of thinking: but runoff elections are often decided by fewer voters, with the winner getting fewer votes than loser got in the primary election. First of all, this probably assumes relative inconvenience for voting, thus lower turnout, in the runoff. In Cary, the reverse was sometimes true, and generally runoff elections enjoyed about the same turnout. But, it is argued, by compressing all voting into one session, it is more convenient for voters, and therefore IRV is more democratic. This is actually a far more complex debate than one might think. Runoff elections probably choose more satisfactory winners, overall, due to differential turnout. Essentially, runoffs test preference strength, the voters who don't turn out for the runoff *don't care as much about the results* as do those who turn out -- generally. Obviously, there are plenty of individual exceptions, such as parents who can't find child care, etc. But these considerations can apply to both elections, and runoffs allow voters who *do* care an additional opportunity to vote. As anyone familiar with Range voting would recognize, this is a true test of sincere preference strength, on average, thus votes are, as with Range, deweighted according to such strength, and it's easy to show, mathematically, that this should result in higher average satisfaction with the result. If the top two are equally acceptable to me, as an example, I may not bother to vote, and, in fact, if that's the only issue, probably won't. So turnout can be lower, and possibly a lot lower, even if there is no differential barrier to voting.
Point: many elections do require a majority for victory. Some of these have election/runoff schedules that may actually favor the runoffs. (Think about the current overall Presidential primary process as the "primary," with the top two facing each other in the general election. This is *roughly* what we have. When IRV replaces a majority-required runoff, depending on how the replacement bylaw is worded, majority might be continued to be required (the Vermont legislation proposed), or not (the San Francisco proposition as actually worded, not as advertised).
Now, as to balance. Richie's edit summary was: "under "practial implications", doesn't seem useful to describe in laborious detail a proposal that no governmental election uses."
This article, first of all, is about voting systems, and not only about those used for governmental purposes. Secondly, the detail given is very brief. Such legislation is pending and, indeed, should be covered by the article (or at least the implementation article).
We have seen this "laborious detail" argument many times. I will refrain from bringing in all the political implications, what I know about Richie's arguments elsewhere and how this dovetails with them, but it is important to note here that, obviously, User:RRichie is Rob Richie, Executive Director of FairVote, the main advocate of IRV. He was here before as a tendentious IP editor, actively removing contributions from "critics," -- yes, that was the edit summary, "editor is a critic" -- and was blocked as such. He is quite welcome to participate here, as far as I am concerned, for he has great and even unique knowledge of the subject. For example, we have a report as to how the term "instant runoff voting" came to be used, but it would certainly be of interest for the article what Richie recalls of that. (Never mind for the moment how we would find reliable source, suffice it to say that there would be a way.) However, he is a COI editor, even more so than User:Tbouricius and, as such, should refrain from any sort of editing that could reasonably be seen as promoting a point of view where he has a conflict.
The judgment of what is "useful" depends highly upon POV. Useful for what? Richie is a political activist, making his living at influencing opinion. Any skillful activist knows that what is not said can be as important as what is said. If one can place before decision-makers a subset of information, chosen to further only one point of view, one can influence them toward that point of view. Other information certainly won't be seen as useful! What is important for readers to know? The fact that IRV disregards, as commonly implemented, inconvenient votes looks bad. But if we can blame it on the voters, by how we state the fact, i.e., with language like "voters who did not bother to rank enough candidates" or the equivalent, i.e., "failed to rank" has been used in the article (and might still be there), one can divert attention from the fact that an election was held, a majority of voters voted for someone other than the IRV winner and did not vote for the IRV winner, but the IRV winner still wins. One may certainly then argue that for various reasons the IRV winner was still legitimate, but these same arguments start to resemble the arguments for plurality. Why do jurisdictions bother to require a majority? Are they aware that they are waiving the requirement if they implement IRV as generally proposed? In Vermont, quite simply, the requirement could not be waived, it is constitutional. *That is the only difference.* This is actually a *crucial* fact. Now, does this have POV implications? Certainly! That is, there is a clear POV motive that one might have for inserting this into the article.
"It will confuse the readers." Sure it will. Without it, they will be quite clear, look at this, all the evidence supports it, etc. Add some more facts, they become "confused." I.e., realize, more, the complexities of the situation and no longer find it so obvious. And out of this confusion can come true clarity. Wikipedia policy trusts the readers to sift through the evidence, but if the evidence is differentially presented, the results can be warped. Wikipedia actually depends on editors with POVs, they do, in fact, most of the work. Someone without a POV on IRV probably hasn't been paying attention! If one has a POV, one can become very sensitive to opposing POV, that might otherwise pass unnoticed. For easily more than a year, I assumed that FairVote was right: "Roberts Rules recommends IRV." It was only my POV that caused me to look more closely, why would this very respected source for parliamentary procedure make such a mistake? Turns out, they didn't make a mistake. RR doesn't "recommend" IRV -- and, of course, this has filled the Wikipedia drives with many kilo-characters of discussion which do not need to be repeated. And, again and again, pro-IRV editors would take out the necessary detail, put in to make the mention of IRV in Robert's Rules accurate and neutrally presented, not excluding contrary detail.
Because of the COI, Mr. Richie should refrain from making edits which could reasonably be expected to be controversial. If he wants to make a controversial edit, he should instead, discuss it here to seek consensus, and not be the one to actually make the edit (unless no controversy actually appears). For details and confirmation of this, see WP:COI. And, in spite of all this, once again, welcome, Mr. Richie. I assume we can cooperate toward the improvement of what has become a family of articles, and my goal has been to help raise this article to Featured Article status. That will require, however, thoroughly cleaning up any remaining POV imbalance (in any direction). As Mr. Richie knows, I generally oppose IRV as an inferior method, but it is quite reasonable for some jurisdictions to try it, and my opposition was originally based almost exclusively on "it is a good reform, but we can do better at lower cost." As I found with some of those who came before me (i.e., people who have been studying elections methods since before IRV was named), I've actually come to oppose it for various reasons, mostly that it doesn't bring sufficient improvement, in practice, to be worth the cost, or, at least, my support has become much, much weaker, more easily influenced by cost. It does not escape my notice, though, that the 2000 election would have had a more democratic outcome if, say, IRV had been in place in Florida for presidential elections. But the same is true for the much, much simpler reform of Approval voting, cost-free, or the alternate "instant runoff" of Bucklin voting, which has the advantage of clearly satisfying the majority criterion, but which is also more efficient at finding majority winners -- it does not miss any votes from the counted rounds. (I need to put the info back in, it looks like I failed to save an edit, but, as I recall, Bucklin analysis applied to the San Francisco election data did find a majority winner in almost all the elections where IRV failed to do so. Bucklin, which requires no changes to equipment or significant change to procedures (Just Count the Votes), really deserves more attention.) -- Abd ( talk) 18:35, 1 February 2008 (UTC)
(unindent)First of all, my thanks to Mr. Richie for taking the time to copy the material from the Encyclopedia Brittanica. It will be interesting to examine it from the perspective of this problem. As to his having more knowledge than I about instant runoff voting, of course he does. He's been promoting it for more than a decade and has created many of the arguments advanced for it. He is not, however, an expert on Voting systems in general, that is very clear from much of what he has written. His vision is narrow. Nevertheless, it is very welcome here. Contrary to what it sometimes seems, expertise is not a disqualification from editing Wikipedia. But experts often don't have the patience necessary. It's a problem, actually.
However, as to Wikipedia process, I could properly be his guide, even though it can be truly arcane and unpredictable. If we can find consensus here, we'll quite likely end up with a better article than if we try to take issues to mediation or arbitration. Now, Wikipedia does *not* expect editors not to make mistakes. We can go out and find statistics and report them. If we find a mass of statistics, we may very well make mistakes. In fact, I expect it for myself. Wikipedia is a collaborative effort, far more than traditional publishing. If I had submitted those San Francisco statistics for formal publication, you can be sure they would have been checked and double-checked before the manuscript went off. Here, I wrote the Talk page edit as I collected the data from the source given, so anyone could check it. In this case, there was an interpretation error, but I'm sure it would have been caught in any normal editing review. None of this should change that was reported was still *substantially* correct. IRV is not performing as I would have expected, nor as it was promoted. It is avoiding runoffs, but the same results would have obtained, in all 33 elections, if a majority requirement had simply and directly been discarded, as it was discarded by implementing IRV, but instead going to ordinary Plurality voting(or, better, Approval voting, or, even better from some points of view, Bucklin voting. The same result would have been obtained if all the elections had been Bucklin, except that all but one election would actually have found a majority -- if we assume that voters had voted the same ballots, and there really is little reason to expect that voters would have added fewer second and third rank votes, Richie's argument -- that I'd expect him to make -- on that is weak. Voters aren't going to consider, most of them, later-no-harm, which is of rare consequence in real elections.
IRV failed to find a "majority of the votes" in seven out of nine elections going to instant runoff, as I recall the statistic. One of the interesting aspects of Brown v. Smallwood is that the court explicitly -- and correctly, based on precedent -- associated votes with "men," not with marks on paper. (Times have changed!). Thus votes really means a majority of voters participating in an election.
Mr. Richie has misrepresented my arguments about IRV and majority. IRV with a majority requirement is a different than IRV without it. I would argue, in fact, that it's a better system with a majority requirement, more democratic. IRV does *not*, in one round, reliably find a majority winner, under all legal definitions of majority, including those of Robert's Rules. Implementing single-ballot preferential voting with IRV or any other preferential method does *not* guarantee a majority, unless you force voters to rank all candidates, which would go over in the U.S. like a lead balloon. So implementing IRV by legislation superseding majority-required legislation is *removing* the majority requirement, not merely finding a cheaper way to satisfy it.
(But I would have thought it would be more successful at satisfying it, I *do* consider that if a majority of votes, after transfers but still counting all the votes as the basis for majority, are for a candidate, this is satisfying the majority requirement. I don't like some aspects of it, the candidate eliminations, which is why I'd favor other ways of analyzing a preferential ballot, but the fact remains that, by voluntarily ranking candidates, the voter has voted for them, and thus may count as having voted for the winner. Coercing the voters into ranking all candidates is not obtaining the free consent of the voters to a result. Bad idea.)
There is nothing in the Brittanica quotation that contradicts this; on the contrary, the Brittanica article confirms the point. One might notice this: "This redistributive process is repeated until one candidate has collected a majority of the votes." This is in reference to Australia, where compulsory full ranking causes the process to succeed in this. There is a difference between "majority of the votes," which language is present in the Vermont constitution, I think, and in any case which refers to all valid votes. Votes which have been substituted are valid votes, and remain such. Exhausted ballots do not thereby become non-votes. They are still votes, simply for none of the remaining candidates. With the Mayor of London, if you didn't vote for the top two, in your main vote and your supplementary vote, you are simply out of luck. Your vote is disregarded, and a "majority" is cobbed together only by discarding your vote quite the same as if you had cast a blank ballot. But you did not cast a blank ballot, and perhaps you even voted for a candidate on the ballot. "Majority of the vote" is a concept which has been used for centuries in deliberative process, and if we are going to discard it, it should be deliberately done, not covered up by pretending that we get a "majority of the vote" by discarding some of the legitimate votes.
Note, however, that in partisan elections, it is quite likely that most voters under suppementary vote would, indeed, vote for one of the major candidates in first or second preference, so this system would almost always find a majority, only in very close elections would it not, sometimes.
It's simple, here. If the fact can be verified, and it is notable, it belongs in the encyclopedia, somewhere, and something directly contradicting it can't be maintained, anywhere. (But if there is controversy, then the sides may be reported as such, and then relative notability can become an issue.) It may even be non-notable, but if it can be verified, a clear contradiction cannot stand. If some source adds 2 objects and objects 2 and reports a total of 5, and it happens there is no other reliable source on the sum of the objects being counted, we don't put "2 of these and 2 more makes 5." We fix the error (and we might note it as such). When sources write "majority," in regard to IRV, they are often simply talking about "majority of the votes remaining after exhausted ballots are set aside." Note that it has been a struggle to even get this into the article, what with the barrage of sock puppets and POV edits.
This article, I must remind Mr. Richie, is not just about IRV in public governmental elections. It is about the *method*, which is proposed for use, sometimes, in organizational elections, the sum of which probably exceed public elections by some measure. (But not necessarily in terms of numbers participating each year.) Further, various organizations, including governments and political activists, are considering reforms, and thus need to understand the *possibilities*, not merely what already is in common use. Wikipedia rules restrict us from putting much here about what is well-known about election methods, because much of it is known from peer discussion, outside of the formal literature. Further, even the formal literature is only accessible to a few.
To bring this back to the point, IRV, when used to replace top-two runoff, not only may fail to find a majority winner, it seems that, in the U.S., it usually does, from the examples before us. (That is, if the election fails to find a majority in the first round, in almost every case, it still fails after vote transfers are completed.) I'm sure this would be of interest to those considering IRV for such replacement. Is it important that a majority of those voting have approved the winner? If so, IRV isn't a full solution. Indeed, no single-ballot election method is a full solution, covering all possibilities, unless we coerce voters into ranking all candidates. Further, it is troubling that a majority of voters -- actually as many as two-thirds of the voters in a totally extreme case -- can vote *against* a candidate in favor of another, and yet that supermajority-rejected candidate can win with IRV. Unlikely to be that extreme, yes, very. But the possibility of a majority winner existing who is passed over by the IRV process is much higher. How high is hard to predict, because of the lack of actual ballot data, but center squeeze is quite reasonable under some conditions, where there is a polarized electorate balanced between two candidates and a good compromise candidate, second choice of most people, first choice of less than both of the other two. As Richie knows, there are other preferential voting methods which don't have this problem, including one which is much, much simpler to count, Bucklin.
From the San Francisco data, if the goal is to replace runoffs, Bucklin did find a majority winner -- the same as the IRV winner -- in every contest in 2004 except the very hard case of having 22 candidates (it was just a tad shy). And Bucklin is very, very easy to count. It is also, it can be claimed, a form of "instant runoff voting," but using a different means of estimating who would win the runoff. In most cases the result will be the same, but not in a center squeeze situation, where the compromise winner will likely prevail. In the U.S., we had STV for proportional representation (good method for that), and Bucklin for single-winner, in a number of places, before the backlash took out all but Cambridge. Definitely, Bucklin should be on the table. But much of this is dicta, here. It is mentioned to show the importance of the issue of "majority of the votes." Which is distinct from "majority of the remaining votes. -- Abd ( talk) 22:10, 3 February 2008 (UTC)
Can we add a section under this name and include differences causes by this requirement and stop with the detailed exceptions all over that make the article hard to read and understand? Tom Ruen ( talk) 21:30, 1 February 2008 (UTC)
Well, folks, for a few months I've been noting that MilesAgain was a sock puppet. It's important because even legitimate sock puppets should refrain from contentious editing. I had suspected he was a sock in the same series as User:BenB4 and User:Acct4 who had been part of the cabal of editors "owning" this article; however, he had been more conciliatory, more willing to compromise, and his disruptiveness stopped just short of motivating me to file a WP:Request for checkuser. Well, now we know why he stopped short. I'd have filed it as a suspected User:Nrcprm2026 (James Salsman) sock, as were the others, it would have taken him out, almost certainly, quite a while ago. He's been tagged and blocked entirely without any action on my part. Thought you should know.
Just on a hunch, I googled "James Salsman" "instant runoff voting." Came up with this:"Though we don't even have a Chapter in Santa Clara County, some of our active members there have got IRV on the ballot, as you know. Kudos to Steve Chessin, James Stauffer, James Salsman, and Russ Paielli! (Hope I didn't leave anyone out!) [1]. Distinguished company. This was from 1998, and I recognize three out of four of those names, and I'm sure some of our other editors will as well.
Here is a URL for an open Salsman post on this topic: [2]
Unless it turns out that two checkusers erred in the identification of MilesAgain as another Salsman sock, he lied to us, explicitly.
-- Abd ( talk) 18:47, 3 February 2008 (UTC)
What happens if the two candidates with the smallest number of votes have the same number? Who gets eliminated? Timeshift ( talk) 23:06, 3 February 2008 (UTC)
I moved content before January 23 to a new archive 5, so my scrollbar will function reasonably again, maybe for another 7 days with luck? Tom Ruen ( talk) 23:58, 3 February 2008 (UTC) Talk:Instant-runoff_voting/archive5
Surveying sentences refering majority in the article (by section in bold), here's what I found:
Intro:
Tom: Majority definition okay on "first round". I'd support a change the second part to a logical description:
Abd:need to have the qualification, if it is not already plain from context. Given the context, I fixed the small error in Tom's language (normally that's not a good thing to do in Talk, but I expect he won't mind in this case.)
Tom: I don't think this belongs in the intro, even if its interesting. I'd merge statement into Non-governmental organizations section with RNRO.
Abd: It's been taken out. Note that this is one of the options I proposed long ago (and implemented). However, certain IRV advocates, one of whom we now know to have been James Salsman, insisted it belonged in the introduction, but attempted to get necessary qualification or context removed. It was very clear POV pushing. My "political agenda" would actually favor the complete explanation to be in the introduction ... but I am *not* editing this article to pursue a political agenda beyond one served by a fair and neutral article on the subject.
Counting the votes:
Tom: Seems qualified and clear
Abd: Maybe. I think we should use uniform terminology. "Active" is a synonym, and undefined, though a reader might reasonably infer the meaning. We should also use the language used in legislation and reliable sources, if possible, and not invent synonyms unless there is utterly no doubt about them. I certainly considered "active" close enough not to edit war over! But i don't think it optimal.
Example:
Tom: Properly qualified, but I think its over-kill to focus so much on this fact with a single example.
Abd: From what has become plain, "majority of total ballots" isn't the language to use. It's "majority of the votes," and we can make clear what that means early in the article so that we don't have to keep explaining "majority of the number of ballots containing at least one vote for an eligible candidate," or some other definition probably taken from a legal source (or Robert's Rules), whence it may be sourced.
Effect on parties and candidates:
Tom: This statement is unclear meaning to me - what does absolute majority mean? I can guess, but I'm not even sure this fact is important enough to give here under a Effect on parties and candidates section. Abd: It's an error, I think. But Australian usage may be different. So this may take some thought. I think it means "absolute majority of the voters participating in the election, having cast a valid ballot." That defines a group membership without specifying, beyond the fact of valid participation, how they voted. Because this is Australia, though, "absolute majority" is guaranteed, it's a tautology, there are no exhausted ballots because if a voter truncates, it's not a valid vote. Am I right about this? So this was a POV edit, though it may not have been intended that way. It was an Australian not understanding the context of the debate that was going on in Talk, I'd guess.
Effect on parties and candidates:
Tom: I think I understand this even if ambiguous. It's related to the power of gerrymandering. Your party can get a minority of the votes and a majority of the seats if you can pack the opposite party supporters more densely into a few districts that win by landslides so other districts can lean towards closer elections in favor of your party.
Abd:I'm not sure this should even be in the article. The intention is perhaps what we might call an "IRV majority." If it meant "majority of the voters participating, casting uncoerced votes," then IRV doesn't guarantee that even locally, not to mention nationally. Nor does any other method. (Methods cannot create support! Methods can only discover it if it exists.) This is really a "single-winner" problem, not an IRV problem. If we could get this IRV nonsense out of the way, we could really cooperate on proportional representation, though we might still debate about which method to use. Frankly, though, STV is right up there among the top choices, and is one with a lot of experience.
Majoritarianism and consensus:
Tom: This whole section seems an unrelated mixture of ideas - talking about majority vs PR on one side, theoretical Condorcet comparisons on the other, and explains the definitions of majority.
Tom: I have to think whatever is worth saving here, it needs to be split into each issue:
Abd: if Condorcet ballots may be truncated, or if they otherwise indicate some distinction between Approval and non-Approval, then it is possible to determine if there is a majority or not. Otherwise, what we would have to say is that a majority of those indicating a preference between this candidate and all other candidates preferred the Condorcet winner. There are no "exhausted ballots" in Condorcet. However, there may be ballots that do not participate in a particular pairwise election. Question is, did a majority indicate a preference for that candidate over all others? That can be determined. Condorcet does not require a majority, but Condorcet methods may be used in elections that do require a majority.
Is IRV better than other systems?:
Tom: The first two go together as and come down to differing definitions of majority. The THIRD is something completely different that I'm sure will confuse people, referencing the Condorcet winner without saying this.
Abd: This section is an arguments section, and arguments are presented as they actually are presented in the real world. These should really be quoted and, no, we don't get to change them to make them truer or more false. The "majority" in the last mention is "majority of the votes cast," or, more accurately, of the voters voting and casting valid votes (i.e., votes that would have been counted if for a winner).
Similar systems:
Tom: NOW apparently we have a NEW definition of majority (approval majority?!) which allows multiple votes, multiple majority candidates, but only one winner. This is NOT good as is in my view!
Abd: Its the same definition. A majority of those voting have, in that round, voted for the winner. Yes, with Bucklin, we have to realize that is is possible (though actually quite unlikely, I haven't seen a real Bucklin election which did it) for more than one candidate to gain a majority. The "lower choice candidate can count against your higher candidate" is a POV summary of Bucklin, highly imbalanced. Show an election where this actually happened and the voter would have been upset about it! The definition should be rewritten, it does not actually explain Bucklin counting (but it might continue to). Again, since it is in the negative, the double majority problem *does not exist*. The winner is, as with Approval, is the candidate with the most votes when counting stops, and it stops when a majority is found. That there might be two or twenty is irrelevant. One stops the counting, the winner is the one with the most votes.
Runoff voting:
Tom: Overall majority?!
Abd: Synonym for "majority of votes cast. Totally unnecessary, since there are no eliminations in the first round. If there is no majority in the first round, there is a new election with reduced access to the ballot. Often, in theory, that new election is still open and majority failure could occur, but I think that this is very, very rare.
IRV in a larger runoff process:
Tom: I don't know where to start, but maybe this summary is helpful. Tom Ruen ( talk) 23:36, 3 February 2008 (UTC)
A difficulty in terminology in this article is the use of the word majority can be defined differently:
No election system can guarantee a majority of voters will support a single winner and the only way to make a promise on majority is to redefine it in a way that exclude voters from the count who refuse to offer a vote among the final set of choices.
If an election implementation requires a majority winner, then it must be defined in a way that limits a winner from being declared and leaves the election result open to a further election process.
Some elections, including variations in IRV implementations, may allow voters a chance to offer a protest choice like None of the above that rejects all choices offered which may binding or nonbinding. If nonbinding it has no effect on the election. If it is binding this vote can block a winner from being chosen.
First of all, majority has a clear definition: "more than one-half." What gets unclear is when the word is used without specifying "half of what"? absolute majority, according to the article, means "half of all eligible voters," not merely those who voted, but including those who did not vote. Absolute majorities are required for certain actions (for example, an absolute majority generally may be able to amend bylaws without notice; because absolute majorities can be hard to obtain due to low participation in many organizations, supermajority rules were developed to make it practical to amend bylaws without jumping through such a high hoop: after proper notice, a 2/3 vote may amend bylaws.) However, the term may be used, and I think I've seen it used that way, in an election as "absolute majority of the votes cast;" the "votes cast" being the number of ballots containing a vote for an eligible candidate for the office. IRV may proceed to ignore some ballots with such votes, but, then, it is finding, not a majority of votes, but of a subset of votes, being those remaining after eliminations. It is still using a "majority" measure, but of votes now qualified as remaining after eliminations.
As I've pointed out, if we think that IRV finds a "majority" in all cases, we could just take the election one more round (i.e., eliminate the remaining candidate and transfer votes, if any, for the winner. Indeed, this would be a useful analysis, and it might find a majority, if anyone cared. I'd expect, in San Francisco, that this would have shown a few more votes for the winner. In one election, it might have tipped the balance. So if one actually wants to find a majority with IRV, elimination should be continued until only one candidate remains, not two. However, the point of bringing this up is that, as this point, the candidate remaining has 100% of the last round votes. That's unanimity, if we only talk about uneliminated candidates.
Now, language being what it is, once some special usage is established, ordinarily a word might then be used without qualification. However, here, because of all the issues, it is dangerous (i.e., risks misunderstanding) if the word "majority" is not used precisely, which requires either some special terminology, or that the basis for the majority always be stated. This problem does not disappear by including a special section on it; I suspect that Tom thought it would, and then "majority" could cheerfully be used even where statements would become incorrect if "majority of the votes" is glossed. The reason is that "majority of the votes" is common usage. In a place using runoff elections, that is the language used; no action is complete without a majority of the votes. This is, indeed, very common practice in NGOs, and it's a basic protection against the vagaries of elections with 3 or more candidates. Robert's Rules discourages eliminating *any* candidates, so if no majority *of the votes* is obtained, an entirely new election is held, no candidates are eliminated (but some may and often do drop out, which is one reason why one-vote methods work).
Preferential ballots make it easier to find a majority approving of the result without requiring tactical voting, even with a fair number of candidates, but even with as few as three candidates, as we've seen, it does not necessarily happen. From my perusal of the Cary election, it appears entirely unclear who would have won a runoff election, Maxwell or Franz. It might be predicted from Bucklin analysis, though I'm not sure the data is available. (It was possible in San Francisco because, in 2004, the number of votes in each rank, total, was reported -- this is from optical scan, I think.) But it was complicated by the advice the losing candidate (eliminated before the final round) gave his supporters: write me in at each rank.)
Now, as to what is said about each method, it's POV. With reference to Tom's first draft, above.
Plurality. doesn't require a majority. (But think of it as being rounds. Eliminate the candidate with the lowest vote at each "round," and keep this up until there are two left. The winner is the one with a majority of the votes *remaining.* Of course, we don't think of plurality that way; instead we do batch elimination of all but those at the top, which is only one if there are no ties.)
In two round system, a majority winner is a candidate who has more votes than all others combined in the first round, or more votes than a runner up in the second round. If fewer voters participate in the runoff round, the winner might have less votes in this round than the first round. This is POV, it's an argument comparing two different elections. Two round system requires a majority of votes cast in the first round *or the first round fails*. Then a runoff is held; this is a new election with two candidates eligible to be on the ballot. In fact, write-ins are typically allowed. Because, however, it is considered that at this point enough is enough, I think that a majority is not required. The argument about number of votes in the last round is POV. Sure it's possibly true, but the reverse also can be true. The sock MilesAgain tried to rake me over the coals on that claim, but I'd been looking at Cary, and the runoffs that had been held there. Turnout was typically about the same for the primary and the runoff, and in one runoff the turnout was higher. Consider this: an election where, unexpectedly, a dark horse candidate finishes second, and there is majority failure, so it goes to runoff. What will happen. Suddenly, a whole chunk of the population might see hope where they previously had none, so they are highly motivated to turn out. The dark horse could win. In the San Francisco elections prior to IRV, in the year studied (2000, which was remarkable for a high number of runoffs required -- one of the little hidden pieces of POV imbalance in the article, a very unusual year was reported as if it were typical, though this certainly may have influenced San Francisco voters in 2002), one-third of the runoffs *reversed* the plurality win in the first round. With IRV, no reversals have been seen. In Australia, reversals occur a reasonable percentage of the time. But Australia requires full ranking, plus those are partisan elections. Nonpartisan elections are, I suspect, more likely to see reversals with runoffs, and less likely with IRV, and the effect of voter motivation may be behind it. Eliminating runoffs by using IRV may "improve voter participation," but not the making of decisions by voters sufficiently motivated to vote. In any case, a comment about the justice or injustice of two round systems is inappropriate here in the article. What's a majority is the question this section addresses, and it is not any different than with "plurality." It's a majority of valid votes cast.
In Instant runoff voting a majority winner is a candidate who has more votes in the final round than all other candidates combined. This winner may have a majority of the active votes, but less than a majority of the total ballots because some ballots offered no preferences among the candidates in the final round. Of course, treat this similarly to the treatment just given to two-round system, and we would add, "In IRV systems with limited ranks, which is every one in use in the United States [is this true?], there may be insufficient ranks for voters to be able to vote sincerely and also vote for those candidates." None of this makes the last round plurality a "majority of votes cast." That is, calling this candidate a "majority winner" is misleading.
In a Condorcet method, a winner assumed to be the candidate who can defeat all others head-to-head. Some call this a true majority. This candidate may NOT win an Instant runoff if he is eliminated before the final round. Even this true majority exists, the majority status can be in doubt since a candidate might win all head-to-head comparisons to competitors but fail to win a majority of ballots in all comparisons due to exhausted ballots, just like IRV. Nobody calls the Condorcet winner a "true majority." Besides the bad syntax. That winner is called "the Condorcet winner," if one exists. Condorcet methods satisfy the Majority Criterion (all versions), but truncation ("exhausted ballot" is not relevant to Condorcet methods) does mean that the Condorcet winner might not gain a majority. But it's unlikely. Condorcet methods consider all the votes simultaneously, so it is more capable of finding a majority winner.
In an Approval election, voters are allowed to mark more than one choice and have multiple votes. This means that it is possible for more than one candidate to have votes from a majority of voters, while in single-vote systems, a majority support can only be held by one candidate. That is correct, but, again, not relevant. If Approval is implemented in a jurisdiction with a majority requirement, there may be more than one candidate qualified to be elected by gaining a majority, but the one elected is the one with the most votes. And this candidate, in this event, has gained a majority. That another candidate, or even more than one other, may have gained a majority is moot. It simply means that more than one candidate could have been elected by a majority. I have proposed, actually, that if this double majority occurs, a runoff be held, because it is not clear that a majority of voters preferred one to the other. However, majority requirements exist for Ballot Questions, and if there are multiple conflicting Ballot Questions gaining a majority, the one with the most Yes votes prevails. Approval Voting, actually, with a majority of votes cast required. In this case, the majority of votes cast is the votes cast for each question, not a majority of votes cast for *any* question. (This would apply to Approval elections if they were Yes/No elections, with a majority Yes vs No required for each candidate to be eligible to win.)
No election system can guarantee a majority of voters will support a single winner and the only way to make a promise on majority is to redefine it in a way that exclude voters from the count who refuse to offer a vote among the final set of choices. Well, they do it in Australia. However, other than coercing voters to fully rank all candidates, which is what they do (by considering ballots spoiled if not fully ranked), it's true: there is no way to guarantee a majority in any election where more than two candidates are allowed. Top-two runoff *almost* guarantees a majority, but typically top-two runoff does allow write-in votes, so, theoretically, it is possible for a candidate to have a plurality and not gain a majority, due to write-in votes. My guess is -- I have not checked -- that majority is not required in the second round of top-two.
If an election implementation requires a majority winner, then it must be defined in a way that limits a winner from being declared and leaves the election result open to a further election process. This is a bit confused. It'd put it: "Generally, if law requires a majority winner for an election, the election may fail to find such a winner, requiring further process. The only know ways to avoid this are to limit the candidate set to two, or, with preferential voting and more candidates than two, to require all voters to fully rank all of them."
Some elections, including variations in IRV implementations, may allow voters a chance to offer a protest choice like None of the above that rejects all choices offered which may binding or nonbinding. If nonbinding it has no effect on the election. If it is binding this vote can block a winner from being chosen. While this is certainly interesting to consider, I've never seen such an IRV variation. Has anyone? NOTA, if nonbinding, means nothing, what's the difference between this and a blank vote? If the vote is counted, is it part of the basis for a majority? If so, then it is binding if a majority is required. Now, in most elections here in the U.S., a write-in vote is allowed. One could write in any name, and, as long as this is the name of an eligible candidate (which usually means a resident of the jurisdiction, of the age of majority -- a different meaning entirely! -- and not disqualified for some reason such as citizenship or felony conviction), it serves the NOTA function; if enough voters do this, it would deny a majority to the winner. But this is all theory. Majority NOTA *should* cause election failure in a democracy. But the meaning for an officer election is unclear.
And ... a huge pile of original research, unsourced. Some of it could be sourced, to be sure. Some, quite possibly not.
By the way, I appreciate the tremendous effort Tom has put into trying to clean up this article. Regardless of whether or not I agree with all the changes he has made -- and it will take some time to catch up to his racing ahead -- his work is appreciated. -- Abd ( talk) 06:30, 4 February 2008 (UTC)
(unindent) Thanks, Mr. Bouricius. It's a little offensive, I'll note, to project upon me POV "fixation." But, hey, I've seen worse said around here. I do not have any intention for majority to mean a number as large as possible; rather, I'm focused on it meaning exactly what it has always meant when we say what is in legislation and constitutions as majority of the votes. I used the wrong number in my original presentation of the data from San Francisco. At one point I also incorrectly asserted that blank ballots might be part of a majority basis, but both these were simply errors. I make mistakes. They do not proceed from any intention to deceive or exaggerate. (As to the ballot issue, "majority of the ballots returned" would be the term, and "majority of votes cast" would be something different, quite clearly, since blank ballots don't contain a vote. Further, there is a legal presumption that an invalid vote, not for an eligible candidate, isn't a "vote." It is the same as a blank.
Thanks for the information about Burlington and Takoma Park. Full ranking is, of course, a fish bicycle in Takoma Park, but I think there were more than three candidates in Burlington. IRV may indeed be lumped together by some as a "majoritarian" method, but this may merely reflect two things: first of all, it satisfies all the definitions and proposed definitions of the Majority criterion, and it does seek a majority. However, it does not maximize majority, for the sequential elimination, for reasons Mr. Bouricius surely knows. That is, it is possible that one may derive, from the votes on an IRV ballot, a winner who would be preferred by an actual majority over the IRV winner, and over every other candidate, and who would, indeed, have a majority of ballots containing votes. For example, suppose there is a set of voters and preferences:
Far from being a weird contrived example (though the extremity of it, the almost perfect balance, is indeed chosen for purpose of illustration), this example represents a polarized electorate, with society divided almost exactly in two (51:49) over some issue or constellation of issues. A represents, perhaps, the middle of the left, C represents the middle of the right (perhaps they were chosen by the Left and Right parties, and B is a compromise candidate, as Robert's Rules surely considers important. B is in the center. So half of B's supporters come from the Left and half come from the Right. If B is eliminated, A wins, narrowly. But 65% of the voters preferred B to A. This is the classic Center Squeeze. B is the Condorcet winner, but because the B "party" -- if there is a party -- is slightly less popular than the two extreme parties, B loses. It is entirely due to the elimination process. Top-two runoff, of course, has the same problem. This is why Robert's Rules dislikes candidate elimination.
Now, facing this, if some of the C voters realize their predicament, they *might* vote, instead, B>C>A, thus turning the election to B, whom they *greatly* prefer to A. However, Favorite betrayal criterion is not the topic here. Consider how Bucklin handles this election: in my opinion, in Bucklin, the vast majority of voters will vote sincerely; and whether or not truncation will be practiced quite depends, I expect, on preference strength. However, if we assume that some of the A voters and more of the C voters realize the risk of truncating, plus the middle third of the B voters truncate (they don't really care that much who wins, A or C, if B doesn't), We might get
(Much of the information and analysis we have comes from FairVote and, in my opinion, it has not been objectively reported, original sources aren't available for checking, etc. The claim that Bucklin voters did not add additional ranks because they didn't want to hurt their favorite is unsupported, even though that phenomenon can be shown to be, in pure theory, strategically advantageous. In fact, the dilemma only arises in partisan elections where a voter is puzzling over whether to vote for Bush and Gore, or just one of them. For third party supporters, where their favorite has little or no chance of winning, it is entirely moot; these are the voters who will add ranks, almost certainly, and some major party supporters will also, where they want to send a message to their own party as to which way to lean, and would not be seriously displeased if it *did* turn out that their vote caused the third party candidate to win. But, by definition, *most* voters do not have a strong motive to add ranks, and that is harmless. But, of course, it *could* cause majority failure. But very unlikely to do so. In nonpartisan elections with many candidates, majority failure becomes very possible, even likely, unless there is a runoff reducing the candidate set (or even not: voters change their minds, decide to compromise, and candidates drop out). But it's worth noting that Bucklin analysis showed a far higher expression of a majority result -- though certainly the argument that voters may vote differently with Bucklin should be noted, though I suspect that, in fact, they won't. Not in large numbers, in nonpartisan elections, except for the minority of voters who are truly dedicated to one candidate. Lots of voters, in the Smallwood elections, did add additional votes, and they turned the result, hence the suit by a supporter of a loser, who had enjoyed a plurality in the first round.)
Now, the situation above produces the same results in IRV as the original fully-ranked results, but, because of the truncations, without a majority, showing something important. The truncations that matter are the B voter truncations, and quite a few of these, actually, can be expected. If we assume a linear issue space, with B in the middle of A and C, voters near B have little predicted preference between A and C, and so may very well not vote for either one; this, indeed, would result from, in Bucklin, a fear that their vote for A or C might cause B to lose, and that is valid, and if they act on it, it may improve the overall result' (from a voter satisfaction standpoint, assuming that relative satisfaction for each voter is as important as that for every voter).
But in Bucklin, because no votes are discarded (unless they are not reached because a winner has been found before that rank is added in), we have no majority in the first round, but A has a plurality. In the second, we end up with A:46, B:53, C:44. B wins with a majority even stronger than the A majority in the original IRV election: The vote counts add up to more than 100, *but* B has a vote from 53% of the ballots, which is entirely accurate. B is a majority winner, and Bucklin has been more efficient at finding that, even in the presence of truncation. What if there is no truncation (which is highly unlikely)?
That's easy. The results are, from the original IRV election, using Bucklin analysis: A:51, B:100, C:49. Truncation does conceal the true support for B, which is actually spectacular. Bucklin can fail to elect the Condorcet winner, to be sure, but it looks like it does so more often than IRV. (There is data on this from Warren Smith and, I think, others, based on simulations). Bucklin finds *every* vote, it does not discard any of them. If voters don't truncate, every voter would vote for B as being in the top two; there might be more candidates in the race, but they were left out for simplicity, these were the rankings of the top three.
So, sure, IRV is considered a "majoritarian" system, but what, exactly, does that mean? What is the definition, what criterion is applied to determine membership in the category? I did not notice either Richie or Bouricius telling us; rather, all I've seen is argument deflecting the point using the fact that "majorit" is found in both words.... and, of course, both are related to "majority" in some unspecified way. IRV does find a *kind* of majority, and some forms of IRV always find that kind of majority: full ranking does it. But IRV does *not* find -- and does not do a good job of seeking, because of vote dropping -- a "majority" as a "majority of votes cast" as it has always been defined, and as it is defined in Vermont. In future campaigns to implement IRV, I predict, this fact will not escape notice if the jurisdiction is using top-two runoff, because those jurisdictions would probably not be using top-two runoff if they did not value verifying a majority of votes cast, and I don't think arguments based on it will be entirely missing from the voter information generally sent to voters, nor will the proposition summary be misleading as it was in San Francisco. So San Francisco might not be repeated, and we may be seeing the peak of IRV adoptions. It is also going to be, I predict, more difficult to find the funding for implementation, if inconvenient questions start to be asked about these issues. So there is my political punditry, such as it is. I've often been wrong. All I know is that the facts uncovered here have not escaped the notice of people working in the field, there are very active opponents of IRV out there (opposed to it for a host of reasons, sometimes they are strange bedfellows), and I don't think they will be caught sleeping again.
Meanwhile, sourced facts, relevant, are not to be removed from the article by editors with a COI, or even mere POV editors, based on vague claims that it is "too much detail," or spurious or off-the-point claims that it doesn't affect "real elections." IRV is not only about actual elections, it is a method and it is being considered for application in many places, political and nonpolitical. An article about the method should consider relevant aspects of the method, and, last time I noticed, Vermont was a notable state. I suspect that the conflict between IRV and a "majority of votes" is also going to affect many other jurisdictions considering the issue, even if they could constitutionally discard the "majority of votes" standard. The only difference with Vermont is that the IRV promoters -- Mr. Bouricius introduced a version of the legislation (the first?) -- had to respect the difference, they could not pretend that the last round was a "majority of the votes," they knew, presumably, that this did not have a snowball's chance of making it past the courts.
When I originally raised this issue here, Mr. Bouricius ridiculed the distinction I was making based on my understanding of Robert's Rules and thus my interpretation of the passage there describing a form of "preferential voting" that is very similar to IRV -- or even identical. Yet, in fact, he knew very well that there was a basis for my objection. I do, by the way, trust his good faith, I don't think he was deliberatively concealing what he knew. I think that, rather, he is as he was, a politician, who thinks in certain ways, and that kind of thinking is why we restrict what Conflict of Interest editors can do, because that kind of thinking is extremely common with them. It's also the norm, in fact, among POV editors of any kind, and my pots are black just like almost everyone else's. I do, however, understand the difference, at least in theory, between my opinions and the truth and what is NPOV. (Those are three different things, but NPOV is as close as humans, absent direct mystical experience -- debatably --, that humans get to expressing the truth.) To find NPOV, I usually need collaboration with others of differing points of view. And to terminate this fugue, some seriously contentious people have learned to work together by understanding that "For our group purpose, there is but one authority, [the truth as it may express itself] through our group conscience." That's a 1940s term for "consensus."
My goal for this article is that all reasonable people with whatever point of view will say, "Yes, that's accurate" and that they will consider it interesting and informative as well. -- Abd ( talk) 20:57, 4 February 2008 (UTC)
(unindent) Thanks, Tom. I do think there is a way to make the article more clear. A section on the issue of majority and IRV may well be a solution; there are some problems with sourcing and original research, but it's possible those could be overcome. But simply settling on some standard terms, defining those (with source), and then using them uniformly where possible, could be sufficient. The point about all this is that the issue of majority is, in fact, an important issue where IRV is being considered. The major elections in the U.S. have been ones where IRV has been implemented as a replacement for top-two runoff, and there has been a great deal of misunderstanding -- and misrepresentation -- about this. So we should simply be clear: what does IRV do and what does it not do? How is it performing? What can be reported from reliable source? There is also the question of where to put some of this. The San Francisco voter instruction pamphlet language from 2002 is important, probably it should be in the Controversy article. The results in terms of whether or not a majority of the votes was found for the winner are relevant, and those facts are available in official reports. We could more efficiently report a summary of all this, but we will, at this point, have trouble finding one. I intend to fix that, actually, but that will take some time. (It involves peer-reviewed publication.) (And, of course, if I'm the author of a paper, ... up creeps Mr. Conflict-of-Interest, perhaps.)
As to executive summary, well, some good writing has been done when someone asked me for such. I tend to respond to requests, it's part of the disorder I mentioned above. I think that if my hangman forgot how to tie the knot, I'd show him
-- Abd ( talk) 03:30, 5 February 2008 (UTC)
Someone inserted information about a survey of disabled voters as it relates to the change in Scotland to STV. This is not appropriate for this article, fundamentally because Scotland did NOT switch to IRV. The editor stated falsely that STV is the Scottish name for IRV. There were numerous changes enacted...such as from single-member constituencies to multi-member (electing three or four seats per district). The fact that STV uses a ranked ballot is only one of numerous changes, and it is wrong to claim the survey was about about IRV. Tbouricius ( talk) 22:50, 4 February 2008 (UTC)
Yes, I'd agree with Tom's suggestion. At the same time, I must point out that User:Tbouricius has, here, made a contentious edit. He should know that, as a conflict of interest editor, that's not allowed. It's absolutely fine for him to argue his points here, but not to revert other editors. His argument has some merit, but it is by no means conclusive. The actual document referenced specifically mentions the ballot form as being a problem, thus the objection could indeed apply to IRV. I edited the reference to make it clear that this was an STV election, and I thought I put in that it "used the same ballot." And the procedure is the same. These were ballots requiring voters write in numbers. That could be difficult for some disabled voters, more difficult than marking a box, and it creates quite a few opportunities for error, such as writing the same number twice. I'll also note that Bouricius totally ignored Ask10questions' quite on-point evidence that IRV and STV are being considered the same; indeed, didn't we just go through a whole controversy over the introduction, with the sock puppet MilesAgain and Bouricius insisting that IRV uses a "single transferable vote," and, indeed, we all know that IRV is simply single-winner STV. Same ballot, same considerations for the voter, same counting method for finding the first winner. STV only deviates in how it proceeds when then looking for the additional winners. All a voter needs to do, with both methods, normally, is to vote for the favorite candidates, in order of preference. However, the voter now faces some consideration of how many candidates to rank, and what to do with all those candidates. With FPTP, the voter simply ignores unknown candidates. I'll agree, that's fine to do with STV as well, but ... does the voter know this? The example in the voter information image I cited shows full ranking. I can easily see some people getting confused. Is this a reason not to do IRV? Probably not the strongest, but it does point out the need for education, which is an additional expense involved (San Francisco spent quite a bit, it's part of the conversion cost.) -- Abd ( talk) 04:28, 7 February 2008 (UTC)
On the other hand, I thank Mr. Bouricius for inserting the reference for the claim regarding Bucklin strategy. I did not find text for that reference, not yet, and it's getting late, but I did find another historical article on the system being used there.
[11]. And there is quite a bit of talk in this about voter strategy. It's a complex topic, and one that I don't think belongs here, unless it is going to be explored neutrally, which would require explaining *other* ways that Bucklin and IRV differ. The argument about Bucklin encouraging voters to bullet vote is, in my view, like many FairVote arguments, not wrong, but misleading. Voters who have a clear favorite simply vote for their favorite, unless they think that favorite isn't likely to win, in which case they add additional votes. The so-called Burr Dilemma is a dilemma allegedly faced by candidates, in how they would recommend voters vote. It doesn't afflict voters nearly so much and, I suspect, most voters will cheerfully ignore it. And the Burr Dilemma is named after a situation that has been called "Approval voting," but that was simply the invention of the author of that paper. It was not Approval voting.
Alabama, it appears, allowed two votes, a first rank and a second rank, and it's legitimate to call it Bucklin because the second rank votes are added in. And apparently, voters did add them in. From the source I've given:
That's fairly silly, actually. The Bucklin results would separately show the first and second rank votes. The sum of all votes is not a meaningful measure in Bucklin, rather the vote percentage, showing true support, would be in excess of 80 percent, almost certainly. That is, this author misreported the results. (But because they did not demand a majority, the Alabamans may have routinely done the same thing: if no majority in the first round, the second round votes are added in and the winner is the one with the most total votes. "Most total votes" will be the same a "highest percentage," no matter what the basis for calculating it is. But if we want to understand what percentage of voters voted for Black, we won't consider the total vote as the basis, rather we may assume that the sum of all first rank votes shows the number of ballots (I'll neglect the possibility of blank or invalid first rank and then valid second rank.)
Since Black got 80% of first rank votes, he must have received at least that much support from the voters. The *minimum* would be 80%; he would get that if none of the voters who voted the other 20% of ballots voted for him in second rank. Let's suppose that the opponents' supporters all bullet voted. In order to reduce his percentage to 68%, the basis must have increased to 80/68 or 117.6%, so 17/80, or 21%, if Black voters, with Black being a very popular candidate, obviously, nevertheless voted second rank for another. The article elsewhere goes on to note: "Second-choice voting also made it possible for a candidate to win with fewer first-choice votes than an opponent." It considers this a defect, but why it's a defect is unclear. IRV can do the same. But it apparently doesn't do it very often; on the other hand, Brown v. Smallwood was a case where it happened, and Brown was a supporter of the opponent of Smallwood who had been the plurality winner in the first rank votes. Apparently the Minnesota Supreme Court also considered this a defect, and it is the basis, essentially, on which they tossed out Bucklin. I.e., exactly the same thing that IRV does. The argument about second rank votes hurting first rank choices is possibly mentioned in Bucklin, but the bulk of the argument from the court is over this impairment from other voters, that they might turn a first rank plurality into a loss. Which, of course, was the whole point of having a reformed voting system. If it never did, why bother?
Now, if we are going to have what is presented here as an argument against Bucklin, in this article, then we will need to have balance. Is that what these editors want? Because it can certainly be done -- but when I've put such in the article in the past, it was vigorously resisted. Until we have some consensus on this, I'm taking the reference to strategy out, it is a complex and controversial issue; but I'm certainly very interested in seeing that source, and in the history of Bucklin in general. It's the closest system we've had to Approval in public elections in the U.S., as far as I know, at least in the Duluth implementation (which really did allow voting for as many candidates as you wish -- in third rank). It's really a ranked Approval, giving first shot to first rank choices, then expanding until a majority is found -- or it terminates with only a plurality. From what I've seen, it would probably be more efficient than IRV at finding a majority unless voters massively bullet vote, more than they are currently truncating in IRV. And this is a huge and distracting can of worms here, I'm trying to stuff them back in at the moment! -- Abd ( talk) 05:13, 7 February 2008 (UTC)