Talk:Instant-runoff voting/archive1

From Wikipedia, the free encyclopedia

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

Contents

intro

4/28/04: Tom Ruen here. I edited the IRV page today, added a second for special cases including rules to break ties. I'm not sure if it is appropriate for the level of article, but it definitely was a vital point missing from the existing article. That is, tie-breaking consideration is vital, but I also included optional rules for quickly eliminating many weak candidates on the first round. Some may object to this approach, although in reality it is very difficult for different elimination rules to affect the winner.

You can always imagine examples where a weak first round candidate wins, but usually it takes core support AND Compromise appeal to win and few candidates survive merely on compromise appeal.

Examples include:

Round 1) A=38%, BD=29%, CD=16%, D=9%, ED=8%
Round 2) A=38%, BD=29%, D=17%, CD=16%
Round 3) A=38%, D=33%, BD=29%
Round 4) D=62%, A=38%

Here a 4th place, 9% supported candidate wins IRV on a cascade of transfer votes, always avoiding last place.

Imagined elections like this are actually the biggest fear from people happy with plurality, that some voters can offer 2,3,4,5 votes to affect the winner while others only get 1 vote.

The reality is that IRV retains the power of the two-party system to thrive because when there are 3 strong parties, voters can't safely vote for a favorite if that favorite isn't also likely to be ranked highly on votes supporting the candidate they hope to eliminate.

First round rules that immediately eliminate a bunch of weak candidates at once help reduce fears that IRV helps weak parties.

Elections like French presidential election, 2002 are actually great example AGAINST bottom elimination. Imagine the recount issues involved in 15 rounds of elimination! Setting a first round survival threshold simplifies the runoff as well as satisfies supporter of the stronger candidates that they will more likely get their earned chance in the final round.


DanKeshet, your reversion was inappropriate. This article as it stands is still biased towards instant-runoff voting without explaining its disadvantages. I've restored the description of the trade-offs versus runoff voting, but left out the description of the trade-offs versus approval voting, save only the simple observation that it's easier to do tactical in preference versus no-preference (approval).

Those trade-offs will go back in, unless I hear a good reason not to do so soon. If you wish to have a project on voting systems, you should be more willing to edit material rather than delete it or revert it. --142

The reason I reverted it was because you are using non-standard terminology. You describe approval voting as one method of instant runoff voting. This is not how the rest of the world uses the term instant runoff voting. You might say that IRV is a misnomer, but you can't just pretend it means something else. Finally, IRV is not a misnomer and approval voting is not a runoff any more than first-past-the-post is. You just pick whoever has the most votes. There is no eliminating of other candidates. DanKeshet

Which places in the world use the term "instant-runoff voting"? How many of these places actually use the system in more than a minor way? Tannin

All the places listed in the first paragraph use it in a major way. Australia is the best example. DanKeshet
Thanks Dan. My question was about the term, however. I have spent more than 40 years here in Australia (plus a little time in New Zealand), and I can assure you that I have never, ever heard the preferential system described as "instant run-off voting" before - and bear it in mind that we get a great deal of our media content from the UK and the USA. On the face of things, I'd say that a change of the article title is called for, but I ask just in case the "instant run-off" term is actually the more common one in other parts of the world that use the preferential voting system. Tannin 16:16 Jan 17, 2003 (UTC)
Ah, sorry. Similarly, here in Cambridge, MA we use single transferable vote (STV) to elect our city council and it's called "proportional representation" (PR), never STV. However, PR, like "preferential system" is a hopelessly generic name--there are many very different systems of PR and of preference voting. The term IRV is used overwhelmingly here in the states; I'm not sure about in other places. DanKeshet
To add to my comment, I believe that in Ireland they call it "Alternate vote"; I wouldn't mind moving it there, though I see no reason as we have a redirect. I see that it's also called "preference vote": it would be a bad idea to move this page there as "preference vote" is a generic term used to describe many systems. DanKeshet
It is also called "Alternative vote" in the UK by those who understand such things, although many people just call it STV. (For a single-winner election, the two systems are identical). To the best of my knowledge, the term IRV is never used outside the US. The article's primary name should be whatever the most common term for the system is in those English-speaking countries that actually use it, so I defer to the Australians on this. Jonathan 15:04, Nov 8, 2004 (UTC)
IRV is used in California, particularly in San Francisco, though probably soon for all elections within California. The term used in California is IRV, though occasionally the term preference voting may be heard. M Carling 04:39, Nov 25, 2004 (UTC)
This is considered a weakness by the advocates of a more deliberative democracy, who point to the French system of presidential election where such between-round dealings are heavily exploited and useful (they say) to draw together a very factionalized electorate. However, proponents of instant runoff voting point to the 2002 French election as an example of where the French system went badly awry; with the votes divided between sixteen candidates, an extremist candidate strongly opposed by most voters was able to make it to the second round with less than 17% of the vote in the first round.

I removed this, not because I think it's implausible that this argument was made, but becuase this anonymous he-said she-said stuff is not really appropriate for wikipedia. Does anybody have one reproducible reference to one person pointing to the French model as a model of deliberative democracy in a criticism of IRV? DanKeshet

The reason I changed the sentence saying that STV was the multi-winner "equivalent" of IRV is because there's more than one multiple-winner system which reduces to IRV in the single-winner case. The other multiple-winner system I've heard of (and heard advocated) does not have a quota at the top, but simply eliminates people off the bottom until there are n candidates left, for an n-sized district. DanKeshet 18:26 9 Jul 2003 (UTC)

"equivalent" was wrong then, but "similar" seems a bit wishy-washy though - I changed it to explicitly say that the one reduces to the other. Better? Martin
Much. DanKeshet 17:44 10 Jul 2003 (UTC)

an idea for preferential balloting

Moved from WP:VP

I have an idea. first, read the article on Preference_voting before going on. now, here's the idea when its mathematically impossible for someone to win, drop them! for example: <html>

Laura 30 votes
Billy 20 votes
Abu 15 votes
Boris 5 votes
Lee 2 votes
Shelia 1 vote

</html>

now, in this example, even if we drop both Shelia, AND Lee, and even if all of their voters vote for Boris, there is no way that Boris can over-take Abu.

I like this idea, but as far as I know its not the law anywhere, though it should be.

I'd like to add to some pages this idea, but am not sure how to do it. I'd like to name this idea after myself of course ;)

Pellaken 02:40, 6 Feb 2004 (UTC)


oh, please respond in my talk page if you have anything 'major' to say, as I am liable to forget about this

They tried that. It was called the 1984 US presidential election. It was a disaster. The media called the election before the polls closed in California. When everyone found out the presidental election was decided, they figured it wasn't worth it to vote. The other California elections got insanly low voter turnout, and a lot of upsets occured. That's why the media doesn't call it before 8 PM pacific time anymore. →Raul654 02:56, Feb 6, 2004 (UTC)
What does this have to do with Wikipedia? -- Tim Starling 03:02, Feb 6, 2004 (UTC)
I think he's asking if his original theory can have its own article on wikipedia. Raul says it's not original. The official line on original ideas/theories this sort of thing is that it's not suitable for wikipedia, right? Or I may have misunderstood the question. fabiform | talk 03:11, 6 Feb 2004 (UTC)
Sorry, I missed the part where he said he wanted to add it to Wikipedia. Pellaken: no, you can't add it to Wikipedia, that would break several rules. Andrewa notes a few below, I would add importance tests under the deletion policy. -- Tim Starling 03:20, Feb 6, 2004 (UTC)

Have a look at What Wikipedia entries are not, especially point number nine, Personal essays, and ten, Primary research. But, if on the other hand the system has been tried, then the system used in this trial could be the subject of an article or a section, together with details of how it worked or didn't. Andrewa 03:16, 6 Feb 2004 (UTC)

Ok, let me clarify the above - he is the first person to actually propose doing it intentionally. The 1984 presidental election was an accident. His idea is definietely original and (as such) definetely does not belong here. →Raul654 03:17, Feb 6, 2004 (UTC)

ok... I have NO clue what the hack raul is talking about. this has nothing to do with the american electoral system, less with the electoral college, and much less then anything involving time zones. Either way, I was basacally looking if I could add this, as a logical alternative, to preferential balloting articles. for example: 'some people think blah blah blah' etc. If not that's cool, but I think it deserves a mention. Pellaken 03:59, 6 Feb 2004 (UTC)

OK, now that the discussion has been moved to a slightly more appropriate place, I'll ask: how does your method differ from ordinary instant runoff? If it's mathematically impossible for a candidate to win, they'll be eliminated in instant runoff after a few steps of redistribution. -- Tim Starling 04:09, Feb 6, 2004 (UTC)

it speeds up the process. for example, in countries like russia, in presidential elections you end up with, for example:

A-40%
B-25%
C-15%
D-11%%
E-1%
F-0.5%
G-0.4%%
H-0.2%
I-0.1%

etc... In situations like this, having a cut-off like the one I suggest could speed up the process greatly. In US presidential elections (assuming they move to a system like this, where each state has a preferential ballot) 99% of the time, a 3rd party candidate cannot, even with ALL other votes, overtake one of the two top candidate. so to prevent votes being counted time and time again (for example, if he put Communist first, Socalist second, Green third, and Democrat fourth) you just make a mathematical cut off, and hence, things work better ... no? Pellaken 04:17, 6 Feb 2004 (UTC)

So what you're suggesting is not a voting method, but an algorithm for calculating the results of an instant-runoff vote. How do you know it's not already in use everywhere? -- Tim Starling 04:21, Feb 6, 2004 (UTC)

This is already standard practice in elections counted by the UK Electoral Reform Society. See this page - regulation 3.c.i. The Electoral Reform Society counts elections for a lot of unions and professional associations in the UK, and its rules would almost certainly be adopted if the UK changed to AV (or indeed STV, where they have an equivalent rule) for public elections. I am tempted to add the link to the ERS's set of rules to the main article - although I think there is a clean pdf somewhere that would be better to use. Jonathan 15:04, Nov 8, 2004 (UTC)

Both houses of the Australian Parliament and of all the Australian states and territories use what Australians call preferential voting. I'm not clear whether this is the same as what this article calls IRV. But you will have to list either all of them or none of them, because they all use the same system. (Some use single-member seats and some are elected by PR, but they are all preferential). Adam 05:56, 6 Feb 2004 (UTC)

Agreed, preferential doesn't (often) mean just IRV, it means preferential voting. Pm67nz 06:22, 6 Feb 2004 (UTC)

If you are restricting the term IRV to "preferential voting with a single winner," then in Australia it applies to the federal House of Representatives, the NSW, Victorian, Queensland, Western Australian and Northern Territory Legislative Assemblies, the South Australian House of Assembly and the Victorian and Tasmanian Legislative Councils. All the other houses use multi-member electorates, either state-wide or by region. The Victorian Legislative Council will switch to multi-member seats from the 2006 election. Adam 06:27, 6 Feb 2004 (UTC)

Further: rather than use a mythological election in Tennessee, why not use a real election from Australia? If people are agreeable, I can supply one. Adam 06:36, 6 Feb 2004 (UTC)

I also agree with Tannin that the term "Instant Runoff Voting" is completly unknown in Australia. In fact, I don't think there is anywhere in the world which uses this term to describe its electoral practice. It seems to be a term restricted to American political scientists and election reform enthusiasts. This should be noted. Adam 06:38, 6 Feb 2004 (UTC)

the Tennessee example is simple, and is used in a number of different voting system articles. any change would have to keep that in mind. Pellaken 06:57, 6 Feb 2004 (UTC)

Re: the name change, I would support moving it to the most common unambiguous name throughout the English-speaking world, whatever that is.
AFAIK there isn't one, because most places that use IRV also use PR.STV, and don't feel much need to distinguish between them. Pm67nz 07:23, 6 Feb 2004 (UTC)
To clarify - that would make IRV the most common unambiguous name, even though the term isn't used in the places where the method is used, because the people who don't have IRV talk about it more than the people who do. Pm67nz 07:28, 6 Feb 2004 (UTC)
I didn't add the Tennessee example, but I like it alot because it demonstrates important differences in the outcomes of different voting systems. If there are similar real-world examples from Australia, that would be even better. If the examples from Australia aren't as interesting, then IMHO, we would have to weigh the advantages of having a real-world example vs. the advantages of having a very demonstrative example. DanKeshet 07:04, Feb 6, 2004 (UTC)
I'm another big fan of the Tennessee example. It uses euclidean space instead of some abstract issue space and so asks less of the reader's imagination and concentration. A list of parties and their platforms is very dry by comparison. Adam has used Australian examples on the pages about the Australian electoral systems, and that's probably where they're most useful. Pm67nz 07:23, 6 Feb 2004 (UTC)

Ah, looking back through the talk I see I've undone DanKeshet's correction. Sorry. STV is a terminological minefield. Pm67nz 07:23, 6 Feb 2004 (UTC)

Here is a good example of a complex preference count in a single-member constituency in the Australian House of Representatives. Charles Blunt was the leader of the conservative National Party of Australia in a traditional rural seat. The intervention of the anti-nuclear campiagner Dr Helen Caldicott allowed the Labor Party candidate, Neville Newell, to win the seat despite polling only 27% of the primary vote. Caldicott also had a good chance of winning the seat - if all of Gibbs' preferences had gone to her as directed, she would drawn ahead of Newell and won on his preferences.

FEDERAL ELECTION OF 1990

RICHMOND, NSW                  73,794 enrolled, 70,571 (95.6%) voted
====================================================================
North Coast NSW: Byron Bay, Lismore, Murwillumbah, Tweed Heads
--------------------------------------------------------------------
1987 two-party majority: NPA over ALP 06.6
--------------------------------------------------------------------
Stan Gibbs                      AD       4,346   06.3  (-00.8)
Neville Newell                  ALP     18,423   26.7  (-08.5)
Gavin Baillie                              187   00.3   
Alan Sims                       CTA      1,032   01.5   
Ian Paterson                               445   00.6   
Dudley Leggett                             279   00.4   
Charles Blunt *                 NPA     28,257   40.9  (-10.2)
Dr Helen Caldicott                      16,072   23.3
--------------------------------------------------------------------
1,530 (02.2%) informal                  69,041
--------------------------------------------------------------------
Caldicott was an independent Green candidate.
--------------------------------------------------------------------

2nd count: Baillie's 187 votes distributed
--------------------------------------------------------------------
Gibbs                 34 (18.4)          4,380   06.3
Newell                44 (23.8)         18,467   26.7
Sims                  21 (11.4)          1,053   01.5
Paterson              35 (18.9)            480   00.7
Leggett               15 (08.1)            294   00.4
Blunt *               17 (09.2)         28,274   41.0
Caldicott             19 (10.3)         16,091   23.3
(exhausted             2                     2                     )
--------------------------------------------------------------------
>                    187                69,041
--------------------------------------------------------------------  

3rd count: Leggett's 294 votes distributed
--------------------------------------------------------------------
Gibbs                 40 (13.9)          4,420   06.4
Newell                17 (05.9)         18,484   26.8
Sims                   6 (02.1)          1,059   01.5
Paterson              50 (17.4)            530   00.8
Blunt *               29 (10.1)         28,303   41.0
Caldicott            146 (50.7)         16,237   23.5
(exhausted             6                     8                     )
--------------------------------------------------------------------
>                    294                69,041
--------------------------------------------------------------------

4th count: Paterson's 530 votes distributed
--------------------------------------------------------------------
Gibbs                 84 (16.3)          4,504   06.5
Newell                60 (11.7)         18,544   26.9
Sims                  57 (11.1)          1,116   01.6
Blunt *              113 (21.9)         28,416   41.2
Caldicott            201 (39.0)         16,438   23.8
(exhausted            15                    23                     )
--------------------------------------------------------------------
>                    530                69,041
--------------------------------------------------------------------

5th count: Sims's 1,116 votes distributed
--------------------------------------------------------------------
Gibbs                179 (16.3)          4,683   06.8
Newell               139 (12.6)         18,683   27.1
Blunt *              562 (51.1)         28,978   42.0
Caldicott            220 (20.0)         16,658   24.1
(exhausted            16                    39                     )
--------------------------------------------------------------------
>                  1,116                69,041
--------------------------------------------------------------------

6th count: Gibbs's 4,683 votes distributed
--------------------------------------------------------------------
Newell             1,555 (33.8)         20,238   29.4
Blunt *              800 (17.4)         29,778   43.2
Caldicott          2,245 (48.8)         18,903   27.4
(exhausted            83                   122                     )
--------------------------------------------------------------------
>                  4,683                69,041
--------------------------------------------------------------------

7th count: Caldicott's 18,903 votes distributed
--------------------------------------------------------------------
NEWELL            14,426 (77.4)         34,664   50.5
Blunt *            4,202 (22.6)         33,980   49.5
(exhausted           275                   397                 )
--------------------------------------------------------------------
>                 18,903                69,041   00.5    07.1 to ALP
--------------------------------------------------------------------
> Charles William Blunt (born 1951): Elected 1984 By, 1984, 1987. 
Defeated 1990
Leader of the National Party of Australia to 10 April 1990
> Neville Joseph Newell (born 1952): Elected 1990
Born: 14 October 1952, Sydney
Career: Teacher.
--------------------------------------------------------------------

Adam 11:35, 6 Feb 2004 (UTC)

It is not correct that Nauru uses single-member constituencies. See the results of the last Nauru election at my website. I still think this article is in a muddle about what it is actually describing. Adam 03:42, 7 Feb 2004 (UTC)

I like Adam's example a lot because:

  • it's real
  • it's from a country where the system is actually used.
  • it's complex enough to differentiate single-winner STV (as we seem to be calling it these days) from some systems, like FPTP.

But I also like the current example a lot for the following reasons:

  • Euclidean space vs. issue space (it's obvious to the reader why voters preferred candidates in the order they do)
  • full sincere preferences and full preferences on ballots cast are available (because, of course, they're imaginary). In Adam's example, we of course are missing the sincere preferences (no way we'll ever know those), but we don't even know the full number of ballots cast in what order, so we can't "re-run" this election under a different system.
  • Similarly, example contrasts well with many other systems. Because the outcome varies depending on the system and the amount of tactical voting you suppose the voters engage in, it demonstrates the differences in the systems and the possibilities for tactical voting very well.

I think that it's possible the best thing for us to would be to replace the current example with Adam's example, which is more centered on the voting system as it's used in the world, but also create a page like Single-winner voting system that compares and contrasts the different systems, which focuses on the theoretical properties of voting systems, and uses the Tennesee example or something with similar numbers but different labels. Please don't change the example on this page just yet, though, my opinion isn't firm, and I want to hear what pm67nz and robla have to say, too. DanKeshet 09:01, Feb 9, 2004 (UTC)

Copied from above by Adam Carr: Here is a good example of a complex preference count in a single-member constituency in the Australian House of Representatives. Charles Blunt was the leader of the conservative National Party of Australia in a traditional rural seat. The intervention of the anti-nuclear campiagner Dr Helen Caldicott allowed the Labor Party candidate, Neville Newell, to win the seat despite polling only 27% of the primary vote. Caldicott also had a good chance of winning the seat - if all of Gibbs' preferences had gone to her as directed, she would drawn ahead of Newell and won on his preferences.

Erm, do you know who would've won if all of Gibbs' preferences had gone to caldicott? (This is a serious question, not a rhetorical one: do you have the data on Newell's second-preferences?) Isn't there a large possibility that enough of Newell's second preferences would have gone to Blunt for Blunt to defeat Caldicott? In this case, if Gibbs' supporters had truly supported both Caldicott and Newell to Blunt, they played their hand well to vote for Newell, who was (hypothetically) "more electable" than Caldicott?
Also, to say that Caldicott's intervention allowed Newell to win the seat is only one way of putting it, though perhaps true. (In the US, where there isn't compulsory voting, the green party turnout for president helped the democrats in some congressional races. I don't know how this dynamic works in Australian elections.) But just based on the numbers above, you could also say that Caldicott very nearly cost Newell the election by siphoning votes from him, and could possibly have thrown the election to Blunt, by the analysis above. DanKeshet 20:11, Feb 9, 2004 (UTC)

If all of Gibbs's prefs had gone to Caldicott, Newell would have been eliminated and his prefs would have flowed very tightly to Caldicott and she would have won. Labor has a very disciplined vote and 95% of Labor voters would have followed the how-to-vote card (see Australian electoral system). The Australian Democrats have a less disciplined vote and not enough volunteers to hand out cards at every booth, which is why their prefs leaked, although it is probably true that some Dems chose to preference Newell rather than Caldicott.

It is unlikely that Newell would have won if Caldicott had not stood. She attracted a personal vote that Newell (a rather dull character) would not have managed on his own. She then sent these votes to Newell as preferences. The general view before the election was that she rather than Newell was the real threat to Blunt. Adam 23:38, 9 Feb 2004 (UTC)

Examples include:

Round 1) A=38%, BD=29%, CD=16%, D=9%, ED=8% Round 2) A=38%, BD=29%, D=17%, CD=16% Round 3) A=38%, D=33%, BD=29% Round 4) D=62%, A=38% Here a 4th place, 9% supported candidate wins IRV on a cascade of transfer votes, always avoiding last place.

Imagined elections like this are actually the biggest fear from people happy with plurality, that some voters can offer 2,3,4,5 v-o-t-e-s/transfers to affect the winner while others only get 1 vote

Comment: I have changed the phrase "2,3,4,5 votes" to "2,3,4,5 transfers" to emphasise the point that each vote has the same value, no matter how many times it is transfered. D 9% of the votes have the same value for each as all the other candidates.

If D with only 9% of the "first prefencence" votes can outrun the others "by the 51% winning post", so be it. There will be always be those who derides D win, and cannot see why A should not be annointed winner, after all A was leading at the turn.

211.30.76.7 07:12, 25 Sep 2004 (UTC)

Preferential Voting in AU - terminology.

In Australia, IRV, Ranked Choice Ballots, and Single Transferable Votes go by the name of "Preferential Voting". And there is some terminology to go with it.

  • First preference - the candidate given the number "1".
  • Second preference - the candidate given the number "2".
  • First preferences - number of first preference votes that a candidate recieves.
  • Distribute preferences - distribute votes of the least popular candidates to the next preference.
  • Eliminate - the candidate with the least number of votes is eliminated, and the votes transfered according to the next ranked preferences.
  • After preferences. The number of voters a candidate gets after lower ranked candidate have their references redistributed.
  • Two-party preferred vote - the number of votes received by the two leading candidates after distribution of preferences.
  • Two-candidate preferred vote - a variation of the previous term if either of the candidates do not come from the big two parties/coalitions.
  • Exhausted vote - where the voter does not number all the boxes, and a preference distribution is unable too proceed futher.


  • First Past the Post - some know this term.
  • How to vote card (HTVC) - a brochure handed out by parties at the entrance to polling booths advising supporters of some of their policies, and how best to fill out the ballot paper(s) to support that parties.
  • Preference deals - since even the candidate with the least vote can influence the result, the recommendations on the HTVCs are subject to negotiation with other parties, presumably on a quid quo pro basis. Voters can of course allocate their preferences as they see fit, since all HTVCs are advisory only.

Unknown terms.

Except from what might be known by observing foreign elections in newspapers or on TV, the following terms are not generally known or used.

  • Runoff
  • 2nd Round
  • Primary
  • IRV
  • Instant Runoff Voting

Other Factors

  • Candidates need not reside in seats, there all seats are contested.
  • Since no Preference Vote is a wasted vote, (unlike FPTP) there are unually 3 or 4 local independents in each seat, who take satisfaction if they garner a reasonable number of votes.
  • Since candidates need no reside in the seats, safe seats are often contested by young members of the opposing big party as a "blooding" exercise.
  • Preference voting takes longer to count than FPTP, and this makes Election Night much more interesting, as votes seep in.
  • There is a 2 or 3 hour time difference between the eastern and western states, and the Western State does not like the result to be all over before they finish voting.

211.30.76.7 07:47, 25 Sep 2004 (UTC)

Instant Runoff Voting is a good name for America

"Instant Runoff Voting" is a good name to use in the US for a voting system known elsewhere by other names because it emphasises the key difference between it and one system it is trying to replace.

  • Instant Runoff Voting - several rounds all on one day or on one ballot.
  • "Slow" Runoff Voting - a first round and a second round weeks later.
  • "Delayed" Runoff Voting - aka SRV

It would help the case for IRV, if better name for SRV and DRV could be found.

For countries not familiar with Delayed Runoff Voting, the following names are OK:

  • Preferential Voting - "number the squares in order of your preferences".
  • Alternate Vote
  • STV for single member seats.

Syd1435 03:13, 2004 Oct 4 (UTC)

How can you help your preferred candidate by not voting for him/her?

I don't quite understand how a voter can "help" his/her most preferred candidate by ranking that candidate lower rather than higher. If my understanding of IRV (at least in the Australian context) is correct, the first preference is important because at each count, the lowest-polling candidate remaining is eliminated. Ranking your preferred candidate lower rather than higher only increases the likelihood that he or she will be eliminated. It does, however, mean that you help your less-preferred candidate.

For example, the Liberal and National parties normally exchange presences, as do Labor and the Greens. If my first choice is the Liberal Party, I should give them my first preference, followed by the National Party (assuming I agree with the Liberal Party's "how to vote" instructions). Now, suppose the National Party received more votes than the Liberal Party, which was my first choice. My vote will be transferred to the National Party. I have helped my second choice, not my first choice. If, however, my first choice was the National Party, ranking the Liberals first would make no sense. Voting Liberal first would not help the Nationals - unless the Nationals received more first preference votes than the Liberals. Voting Liberal first would only help the Liberals finish ahead of the Nationals (the party I sincerely wanted to win), with most National votes then flowing the Liberals' way.

Could Adam Carr or some other knowledgeable person look at this article again? I really am perplexed about how ranking your first choice LOWER could make it easier for him/her to be elected than by ranking that candidate higher. Please explain. David Cannon 11:03, 19 Oct 2004 (UTC)

With IRV, you can NEVER help your preferred candidate by voting first for some one else. There is always a risk of your first preferred candidate falling behind and being eliminated. Should that happen, your vote stays alive in the guise of its second preference.

Thus if you vote National, when you really want coalition partner Liberal, you may get a National, but at least you do not spoil things so that you get a dreaded Laborite. It is assumed here that your second preference goes to Liberal after the National.

Other voting systems such as Limited Voting, Approval Voting, Cumulative Voting are less able to prevent a second ranking choice being elected over a first ranking choice.

The assumption made by Davidcannon is false.

Syd1435 06:25, 2004 Oct 23 (UTC)

How to help an Amazing A-candidate by not voting for him

Imagine you have three candidates to choose from

A is Amazing and your favorite.

B is Bad

D is Dreadful

Imagine that a dual between A and D will end in a landslide victory for your favorite Amazing A. Most of those who initially prefer B at least have the sense to rank A ahead of D, and if B is eliminated his or her votes pass over to A.

On the other hand a dual between A and B will be closely contested. Many of those who prefer D will prefer B before A.

Before the first vote A is doing very well in the polls, so well that the only uncertainty is whether B or D will continue to the run-off.

Now on election day...

If you cast your sincere vote (ABD) the dreadful D is eliminated. Unfortunately for you, in the close runoff between A and B you see that B wins...

If you cast this vote instead (DAB) the bad B is eliminated and a runoff between D and A is arranged. A sense of shock and amazement passes through the country that such a dreadful candidate could pass on to the runoff, but the dreadful D is soundly defeated at the runoff. The fact that you put D ahead of A does not matter since the supporters of B give a ton of votes to A. By putting A down a notch you have secured his election.

This tactic works even better if you can change your vote before the runoff putting A ahead of D.

Something like this may have happened in the French presidential election in 2002. There was an instant runoff between the moderate conservative Jacques Chirac and even more right-wing Jean-Marie Le Pen which ended in Chirac winning 80% of the vote. Had the runoff been between Chirac and the moderate left wing Lionel Jospin the runoff would have been a lot closer. Is it possible that some people believing that Chirac was Amazing, Jospin Bad and Le Pen Dreadful voted for Le Pen in order to secure their favorite Chirac an easy opponent in the runoff?

- But the French election doesn't use an *instant* runoff - it was a straight runoff with multiple rounds (as the article makes clear). This technique is much more dangerous in IRV, because if too many people try strategy with D they might accidentally win the second contest as well as the first! -- JoeNotCharles

FD: the french thing was a spoiler effect in the first round. Because the first round has two winners, the spoiler effect can apply to the second favourite candidate by the condorcet criterion.

Summability Criterion

I have deleted the statement "Condorcet methods also comply with the Summability criterion, making them easier to audit" because it is simply not true that the Condorcet criterion implies the Summability criterion. -- Markus Schulze 21 Nov 2004

POV

This comes across as having quite strong POV in favour of Condorcet variants against IRV - and I speak as someone who shares that POV. I've tried not to reinforce that in my edits, but maybe only an IRV supporter can really eliminate that. ciphergoth 15:17, 2004 Nov 21 (UTC)

We don't need to "eliminate POV", at least not in that sense. If anything, we might want to balance the scales by adding POV on the IRV side; we should never try to remove POV on the Condorcet side just to make things NPOV. An encylopedia which steers clear of stating anyone's viewpoints on an issue is useless - a good encyclopedia just presents both sides of the issue. So we need someone to add some supporting articles of IRV, not delete supporting articles of Condorcet. - McCart42 (talk) 04:18, 15 July 2005 (UTC)
I agree. --Hermitage 08:13, 15 July 2005 (UTC)

Wipe-out results with First Past The Post.

One of the supposed advantages of FPTP voting is that it leads to strong governments with healthy majorities.

The page Wipe-out results (elections) is accumulating examples where the FPTP voting system has lead to "Wipe-outs" or "Clean Sweeps" against the opposition.

The best example is a 55-0 victory in New Brunswick, Canada in 1987.

You cannot get stronger government or healthier majority than that.

The good government and accountability rely on a "reasonably strong" opposition presence in parliament, FPTP fails in the second or corollory part of the need-strength for both government and opposition test, which is a good reason for it to be at least partially replaced.

For example, the New Brunswick parliament might have 10% of seats allocated to non-constituency members, chosen from the best 5 non-government runner-up, and these 5 would be allocated resources as if they formed a party albeit of independents, sharing the resources amongst themselves as they see fit. If 10% is too little or too much, change the figure to one that you like better. Alternately, use somekind of "top-up" PR system. Syd1435 02:56, 2004 Nov 23 (UTC)

Proposed revert: Failure to pick good compromises

IRV's failure to pick good compromises is already well illustrated with the Tennessee example. This addition seems to be an expression of the author's personal thoughts on IRV and voting systems, rather than a description of an existing understanding in politicial science. A voting system, not one known to me, is half-described without any detail being worked out. I don't think it adds to the article. I'll revert it shortly unless anyone wants to speak up against it. ciphergoth 15:52, 2004 Nov 23 (UTC)

Can you show where the Tennessee example illustrates that a broadly supported compromise can be skipped? I see nothing resembling it in there. In fact, the whole article does not contain the word 'compromise'. So I disagree that the section is redundant.
As for the rest, perhaps the section was awkwardly written. However, wholesale removal would seem reasonable only if the IRV criticism presented is incorrect, or too obscure for an encyclopedia article. Neither seems to be the case to me.
I will not re-insert the section, pending discussion, but would prefer improvement of the section to total removal. 195.64.88.137 18:18, 23 Nov 2004 (UTC)
Thanks for your reply! AFAICT your criticism comes down to saying that IRV can fail to pick a Condorcet winner where one exists. In your example, C is the Condorcet winner. The same can be seen in Tennesee, where Nashville is the Condorcet winner and the clear best compromise, and Knoxville is pretty much the worst choice possible. It's possible there's a case for making another example, though - I'd be interested to know what other people think. The discussion of what's wrong and how to fix it needs to be written by someone with seriously expert knowledge - Condorcet doesn't use any of the fixes you propose, but makes the right choice in your example. It's certainly possible that the tendency to knock out plausible compromises early should be made more explicit though. I'd really like to hear some other contributors comment on this. ciphergoth 21:04, 2004 Nov 23 (UTC)
Ah, I see and agree. As far as I can make out, there are three main practical 'failure modes' of IRV:
  • Vulnerable to some of the "usual" tactical voting ploys (compromise, burying) (non-IRV-specific)
  • 'Return of the 3rd-party spoiler effect' when the 3rd party reaches a competitive size
  • Sometimes picks a polarized choice over a broad compromise, giving lower real voter satisfaction
I think it would be nice to put these points in 3 or 4 seperate sub-subsections (improves clarity), and make a passing mention to Condorcet. The last point can be written up with some of the ideosyncrasies of the deleted section dropped.
If there is no major objection I could try something tomorrow and let it be edited (or deleted :-) from there. 195.64.88.137 23:17, 23 Nov 2004 (UTC)
Sounds great, go for it!

POV #2

In my opinion, the recent changes of the "Instant-runoff voting" article are a violation of Wikipedia's Neutral point of view policy. Especially, the claim that IRV was vulnerable to burying and that Condorcet reduced this problem is not appropriate since exactly the opposite is true: IRV isn't vulnerable to burying while Condorcet is. See here. Markus Schulze 24 Nov 2004

1. The text about 'burying' has been there at least since May 2004 (I stopped checking further back at that point). So it has nothing to do with "recent changes".
2. The original author seems to have conflated "burying" and "push-over".
3. I do not see how NPOV enters into it that much? It seems to be error instead of bias.
I will try to make a small fix in naming. If someone can add an example, all the better. 195.64.88.137 00:06, 25 Nov 2004 (UTC)
Example: "Flaws of IRV: (...) IRV is unusual in that it does not satisfy the monotonicity criterion. (...) It must be noted that the Condorcet system reduces or avoids these failure modes, without introducing new problems on the same scale." I consider this comment to be bias because of the following reasons. First: It is simply not true that the Condorcet criterion implies Monotonicity. (You should always make clear when you talk about a concrete Condorcet method and when you talk about Condorcet methods in general.) Second: Whether Condorcet introduces new problems is rather subjective; the Condorcet criterion is incompatible with later-no-harm, later-no-help, and invulnerability to burying, while IRV satisfies these criteria. Markus Schulze 25 Nov 2004
Well, this is a different issue than your original point on burying (thanks for catching that, by the way).
  • Yes, "Condorcet" should refer to "Condorcet methods".
  • The claim meant to be made in the article is that the better Condorcet methods have fewer severe problem situations than IRV. That is the consensus impression I got from the literature. Can you provide a reference to a primary source where the reverse claim is made (i.e. that averaged over all possible scenarios, the outcome of IRV is equal to/better than the Condorcet methods)?
  • If there is no such primary source, but you nevertheless feel that the reference to Condorcet is not neutral enough (downplays Condorcet problems), you could add some text to that effect. Although detailed discussions should perhaps belong in the Condorcet article or in a new "IRV versus Condorcet" article. 195.64.88.137 15:28, 25 Nov 2004 (UTC)
You wrote: "The better Condorcet methods have fewer severe problem situations than IRV. That is the consensus impression I got from the literature." Well, the main problem is that most people read papers only of authors who are known by this reader to share this reader's opinion. I suggest that you should also read papers by Bartholdi and Woodall. Anyway, the Wikipedia pages should contain information about the different Voting systems rather than opinions about the different Voting systems. The claim that Condorcet "reduces or avoids these failure modes, without introducing new problems on the same scale" or that Condorcet "has fewer severe problem situations than IRV" is an opinion and not a fact. Markus Schulze 25 Nov 2004
Alright, you claim that there exists no such rigid consensus (although I note that the 2 papers you referred do not directly contradict the specific claim made in the Wikipedia article, but let's not split hairs too much). What would you propose as an alternative text for the article? Is the text
"It should be noted that Condorcet methods avoid the Return of the '3rd-party spoiler effect' and Failure to pick a good compromise problems"
better acceptable?
Purely as a aside ("meta-discussion"): you seem to feel that Wikipedia has a method of distinguishing "fact" or "pure truth" from "opinion" and then recording only the former. That idea is pretty outdated. The only thing that Wikipedia can do is record the "consensus opinion shared by virtually everybody" without a qualifier, and present the other non-consensus opinions evenhandedly. You'll find a normal article on the Apollo moon landings etc., despite (a very few) heated differing opinions. The problem with the disputed "Condorcet vs. IRV" text is not whether or not it is an opinion, but that it turns out to be less than "almost universal consensus" opinion.
It seems to me that you don't know what an opinion is and what a fact is. "IRV violates Condorcet and Monotonicity" is a fact while "Condorcet reduces or avoids failure modes, without introducing new problems" is an opinion.
By the way, if you really believe that there is a "consensus opinion shared by virtually everybody" that IRV was better than Condorcet then why are there so many IRV supporters? Are they dishonest or only too stupid to read scientific papers? (By the way: Bartholdi's paper is about IRV; he only uses the term "single transferable vote" for IRV. Also Woodall's paper lists many problems of Condorcet; he writes: "An election rule that satisfies Condorcet cannot satisfy any of the seven properties crossed in the column headed 2 in Table 1: participation, mono-raise-random, mono-sub-top, mono-raise-delete, mono-sub-plump, later-no-help, later-no-harm.")
You asked: "What would you propose as an alternative text for the article? Is the text "It should be noted that Condorcet methods avoid the Return of the '3rd-party spoiler effect' and Failure to pick a good compromise problems" better acceptable?". I suggest: "IRV violates local independence from irrelevant alternatives." Markus Schulze 26 Nov 2004
  • Your suggested text is not so much an improvement of the disputed text as it is different information?
  • I think the external links you provide would be very useful to add to the primary sources of the article.
  • "if you really believe that there is a consensus opinion" Lord, I have already granted that it appears that that belief was incorrect. You're beating a dead horse here?
  • "you don't know what an opinion is and what a fact is." You're entitled to your opinion. :-) Your example is right: there is one narrow context which allows for unambiguous facts, formally defined systems (mathematics), where you can work your way from axioms to a result and thus make definite statements. (such as Condorcet/monotonicity criterion compliance). So, you're completely on the mark on that point. However, outside the rigidly defined-in-advance field of mathematics, pretty much any statement is an opinion, even if pretty much everyone accepts it.
Well, if you can say something unambiguously then you should say it unambiguously. Why should someone say that IRV "fails to pick a good compromise" when he could simply say that IRV violates the Condorcet criterion?
You wrote: "Outside the rigidly defined-in-advance field of mathematics, pretty much any statement is an opinion, even if pretty much everyone accepts it." Nope! Examples: "IRV violates Condorcet and Monotonicity" is a fact and not an opinion. "The Condorcet criterion is incompatible with participation" is a fact and not an opinion. Markus Schulze 26 Nov 2004
Why should someone say that IRV "fails to pick a good compromise" when he could simply say that IRV violates the Condorcet criterion? Because the former is immediately understandable to any lay reader, and the second is not. Wikipedia is a general encyclopedia where it is a bonus to use plain language where reasonable.
As for the "facts vs. opinions": you do not provide any new points in your last entry. All your examples (Condorcet criterion, Monoticity criterion, Participation criterion) are concepts with a rigid definition in a formal system (i.e., they are mathematics). So they do not work as counterexamples against the "Outside the rigidly[...]" claim that you quote.
Well, the claim that "IRV fails to pick a good compromise" is true only when you add that with a "good compromise" you mean a Condorcet candidate. Otherwise, a "good compromise" can mean almost everything.
You wrote: "All your examples are concepts with a rigid definition in a formal system (i.e., they are mathematics). So they do not work as counterexamples against the "Outside the rigidly[...]" claim that you quote." So you say that any statement is either ambiguous or mathematical? Then I recommend that you should formulate everything in a mathematical manner. Markus Schulze 26 Nov 2004

Formatting

I removed the italics from the last two paragraphs of the "Effect on parties and candidates" section. Is there a reason that these two paragraphs were italicized? StuTheSheep 20:26, Mar 9, 2005 (UTC)

Fahrenheit's changes

I've just reverted most of Fahrenheit451's changes, which I think are inaccurate and POV. First of all, a quick bit of research would've confirmed the information about IRV in student elections[1]. Also, IRV does not meet "none" of the formal voting systems criterion: as I understand the Arrow theorem, IRV conforms to both unrestricted domain and non-imposition.

Then it should be specified that irv only meets two criteria. Few is inaccurate. I will amend the article accordingly.--Fahrenheit451 18:17, 29 Mar 2005 (UTC)

Finally, there's sentence:

However, such protest votes have an unpredictable effect on an election outcome as IRV fails to comply with any established election method criteria.

In addition to the election method criteria, the accuracy of this statement is highly dubious. It's true that IRV can result in some wacky results under certain mathematically possible situations, but in practice (where there are very few major candidates, and frequently only two), the result is fairly predictable -- protest votes for minor candidates cease to have an impact on the ultimate result.

RadicalSubversiv E 01:39, 29 Mar 2005 (UTC)

Then if there are only two candidates running, irv is redundant and majority vote would apply. How fairly is "fairly predictable"? When rating a candidate lower can cause them to rise in the tallies or when a second choice candidate who would be acceptable to those who voted for rivals in a three candidate election can be eliminated, this election method can regularly produce, as you say, wacky results.--Fahrenheit451 18:17, 29 Mar 2005 (UTC)

It appears to me that Radicalsubversiv does not know the difference between research and counting. The former involves a diligent, in-depth study, while the latter only observation and basic skills of arithmetic. He also seems to confuse established voting method criteria with attributes in Arrow's theorem.--Fahrenheit451 03:14, 25 Apr 2005 (UTC)

"The phrase 'original research' in this context refers to untested theories; data, statements, concepts and ideas that have not been published in a reputable publication ..." (WP:NOR, emphasis added). Please supply a citation to a "reputable publication." There is absolutely no guarantee that the names listed for voting systems in Wikipedia articles are exhaustive. RadicalSubversiv E 04:11, 25 Apr 2005 (UTC)

Proposal to consolidate STV and IRV articles

I propose that the IRV article be consolidated into the STV article with a IRV redirect page. Both terms are often used interchangabily, but irv is specifically the single winner version of STV. I request comments to this change.--Fahrenheit451 21:41, 25 May 2005 (UTC)

It's true that IRV can be thought of as a special case of STV, but I don't think that they should form a single article. A few reasons:

1. IRV does not have to deal with quotas or surplus transfers... thus it omits many of STV's rules.

2. IRV is used for a very different purpose than STV (majority rule vs. proportional representation).

3. Single winner systems are evaluated with respect to criteria that do not apply to multiple winner systems.

4. IRV and STV have different strategic concerns.

5. Both articles are already complex enough that they are somewhat difficult to manage. (E.g. STV has an awkward sub-article entitled "counting single transferable votes".)

I think that there should be prominent links between the two articles that make the relationship clear, but that forming a single article would be confusing and unwieldy. Hermitage 01:04, 9 Jun 2005 (UTC)

By wikipedia standards, my above stated proposal is stale. On the discussion page of STV there was sufficient objection to the proposal to abandon it. Your objections represent a cross-section of what was stated there.--Fahrenheit451 18:51, 10 Jun 2005 (UTC)

Okay; I didn't know that. Hermitage 00:14, 11 Jun 2005 (UTC)

Complying criteria of irv

Someone edited the article to state that irv passed the "pareto criterion'. This is trivial as all voting methods that are used are pareto optimal. For this reason, it was deleted. The 'later no help" and "later no harm" criteria are not published, peer-reviewed criteria, so have been deleted from the article. --Fahrenheit451 01:16, 31 May 2005 (UTC)

I accept the deletion of pareto, although I do hold that approval voting (with a random tiebreaker) fails Pareto, which contradicts your claim that all voting methods that are used pass it. The LNH criteria have been published in journal articles by Woodall. I don't know what makes you believe otherwise. However, I find it more useful to simply say that IRV is immune to burying. Hermitage 00:47, 9 Jun 2005 (UTC)

Why do you feel it is more useful to say that IRV is "immune to burying"? It seems to me that this is no different from making a new criterion. One has to know what "burying" means. It's not obvious to me that "immunity to burial" is the same as satisfying both LNH criteria. I think it would be better to use a rewording of SPC. KVenzke 04:28, Jun 12, 2005 (UTC)
KVenzke, in adding that semi-criterion, I was emulating the Shulze method page, which lists "invulnerability to burying" and "invulnerability to compromising" under the failed criteria, with links to the tactical voting article. I'd be happy to say instead that it passes SPC, but SPC is not defined in Wikipedia. Hence the reference to the tactical voting article. Hermitage 09:28, 12 Jun 2005 (UTC)

You are correct on the point of the random tie-breaker for Approval violating Pareto. However, my assertion applied to typical applications, not "drawing straws". Also, we need to be consistent in how we use voting method criteria for conformity on each listed method. We cannot use the commonly accepted criteria for one method, then use the lesser known criteria for others.--Fahrenheit451 18:56, 10 Jun 2005 (UTC)

IRV is not a Condorcet method, so evaluating irv with against the Smith set is incorrectly included. Stating that a method is not Condorcet compliant implies it is not Smith or Schwartz compliant as well. We need to establish a proper consensus on what criteria we evaluate with and when. That should prevent some editing wars!--Fahrenheit451 00:49, 15 Jun 2005 (UTC)

POV #3

I feel the need to add an npov tag. This article really does have a blatant anti-IRV slant.

I personally have plenty of doubts about IRV, in that I doubt that it can deliver what some of its supporters promise. I do think that it is important for strong critiques of IRV to be available, and for more people to be aware of the alternatives...

But this seems excessive. Critiques should be identified as critiques, and phrased in a more objective manner when they deal with value judgements and make conclusions that cannot be proven, e.g. "some argue that...", etc. I'd like to restore neutrality myself, but it looks like it will take more time than I'd like to spend at the moment. Hermitage 00:46, 9 Jun 2005 (UTC)

Ach... it's hard for me to tear myself away from this stuff. I went ahead and spent who knows how much time working on this article, mostly in the "assessing IRV" category. Does it seem better to anyone, now? Should I remove the tag? Hermitage 01:56, 9 Jun 2005 (UTC)

Well, nobody else jumped in, but I made enough changes that I'm now satisfied with the neutrality of the article, and so I've removed that tag for now. Hermitage 02:15, 11 Jun 2005 (UTC)

Comments on changes

Some argue that Condorcet methods and approval voting are better at selecting compromise candidates and at reducing the spoiler effect.

"Condorcet methods" has been cut. I'm not sure why. I'm not sure if this kind of sentence is appropriate at all, though, without some source. KVenzke 00:15, July 12, 2005 (UTC)
I intend to re-insert "Condorcet methods". I think that it is reasonable to call a canidate a "compromise candidate" if that candidate is favored by the median voter, especially if there is a near-normal distribution of political attitudes.
As to whether the sentence is appropriate, I think so. In this page, I think that we should aim for a multiple POV approach where the pro and con arguments are made clearly. That said, I think that the page is probably still too heavily weighted towards the con sign. I tried to make it less strongly anti-IRV a few weeks ago, and I think that I succeeded in part, but I said at the time that there was still more work to be done. --Hermitage 05:29, 12 July 2005 (UTC)

Also, IRV typically does not allow equal rankings, whereas Condorcet and approval do.

"Approval" has been cut. This sentence is odd anyway, isn't it? Condorcet methods need not allow equal ranking. KVenzke 00:15, July 12, 2005 (UTC)
Yes, I agree that the sentence itself is unnecessary. IRV can be done with equal rankings. Condorcet can be done without them. Approval doesn't use rankings at all. --Hermitage 05:29, 12 July 2005 (UTC)

Unlike Condorcet methods, IRV does not reliably choose the option closest to the center of the spectrum.

has been changed to

Unlike monotonic, multichoice methods, IRV does not reliably choose the option closest to the center of the spectrum.

I don't know what "multichoice" means. But it's certainly not the case that being monotonic means a method will "reliably choose the option closest to the center of the spectrum."
To the end of the same paragraph, this is added:

Because IRV fails monotonicity, less favored candidates can rise in the tallies, while highly favored candidates can fall. This is a serious deficiency of IRV.

This first sentence isn't very clear. It seems to be POV to say this is a "serious deficiency" of IRV; not everyone seems to think this. KVenzke 00:15, July 12, 2005 (UTC)
I intend to change this back. Monotonicity doesn't have anything to do with whether a method chooses an option at the center of a spectrum. The name of the heading is "failure to pick a compromise candidate", and discussion of monotonicity belongs under a different heading. --Hermitage 05:29, 12 July 2005 (UTC)

spectrum example for IRV, Borda, Condorcet

I wrote

"Imagine that candidates are located along a one dimensional ideological spectrum, and that the center of the spectrum is defined by the median voter. Unlike Condorcet methods, IRV does not reliably choose the option closest to the center of the spectrum."

F451 changed to

"...Unlike the Borda count or Condorcet methods, IRV does not..."

This seems to be misunderstanding or a factual mistake. Borda count does not necessarily choose the candidate closer to the median voter under these conditions. For example, imagine a spectrum with candidates at the following points: A at 0, B at 12, C at 20, D at 58, and E at 100. Imagine also that there are 15 voters located at 0, 15 voters at 12, 15 voters at 20, 25 voters at 58, and 30 voters at 100. Assuming that voters rank candidates by proximity, the following set of votes results:

15: A>B>C>D>E
15: B>C>A>D>E
15: C>B>D>A>E
25: D>C>E>B>A
30: E>D>C>B>A

The median voters are at point 58, along with candidate D, who is of course the Condorcet winner.

In IRV, the winner is either C or E, depending on which candidate is randomly eliminated at the beginning. (If it is A or C, then E wins; if it is B, then C wins.)

In Borda, the candidate scores are

A: 105
B: 205
C: 270
D: 250
E: 170

The winner is C. --Hermitage 08:42, 14 July 2005 (UTC)

The best example of the failure to pick a compromise is the Tennesee example. I really appreciate the effort of whoever thought that up; it illustrates the methods clearly and without adding any bias. - McCart42 (talk) 00:14, 15 July 2005 (UTC)
Okay, but Borda still picks the middle in that example, right? I was trying to illustrate that Borda doesn't always pick the middle in 1 dimensional space. --Hermitage 08:30, 15 July 2005 (UTC)

And that is arbitrary. Why not choose a candidate closer to the mode voter? Why not chose one closer to the arithmetic mean? Why not chose one closer to the geometric mean? That is POV clearly biased to condorcet methods.--Fahrenheit451 19:59, 24 July 2005 (UTC)

I reverted the deletion of this paragraph. I don't believe that it is POV; as shown in the Tennessee example, Condorcet methods choose the central capital, while IRV chooses one end of the state, and plurality chooses the other end of the state. Very clearly, neither of those methods are good at finding a "middle of the road" candidate in a common example. - McCart42 (talk) 22:51, 24 July 2005 (UTC)

You argument is totally absurd as the Tennessee example is contrived. It does not clearly demonstrate anything. --Fahrenheit451 20:21, 5 September 2005 (UTC)

No F451, not the mode, the median. (The mode could be plurality, perhaps.) If you have a 1 dimensional issue spectrum, the mean is difficult to measure and difficult to define. The median is much easier to define, much easier to measure, and much more stable. --Hermitage 00:02, 25 July 2005 (UTC)

The mode would be plurality in this case. I understand what the median is, but it is arbitrary to state that this point is the "best" choice in 1d space. Borda count is a weighted mean and gives a good choice as well, as you cite, that is C.--Fahrenheit451 00:56, 25 July 2005 (UTC)

The article does not say that the median is the best choice in 1d space. It says that Condorcet methods necessarily choose the median, but IRV does not, and it suggests the median as a possible definition of "compromise candidate". We can leave it up to the reader to decide whether this definition is satisfying, and whether choosing a "compromise candidate" is desirable when "compromise candidate" is so defined.
I've never heard of the concept of the weighted mean of an issue spectrum. Perhaps this is because we do not read the same articles. --Hermitage 01:56, 25 July 2005 (UTC)

Here is a counter-example to your condorcet POV example:

1,000,000 votes as A>B>C 1,000,000 votes as B>C>A and the last vote as A>B>C. The condorcet winner is A and the Borda winner is B. 1,000,001 voters selected A in first place, but 1,000,000 voters selected A in last place. However, 1,000,000 voters selected B in first place and 1,000,000 voters selected B in second place. It is non-sequitur that A should be the winner for obvious reasons. --Fahrenheit451 20:28, 5 September 2005 (UTC)

Reply to Center for Voting and Democracy's IRV over AV stance

This post from the Election-methods mailing list in October 2004 is a response to the Center for Voting and Democracy's letter in support of instant-runoff voting over approval voting. It contains many good points on the nature of fairness in electoral methods. - McCart42 (talk) 04:14, 15 July 2005 (UTC)

Yes, that's a good letter. My personal opinion is that approval and IRV are both worth supporting. Since most elections in the USA use plurality (some use runoff), and different places can change to different systems, supporters of IRV and approval shouldn't see themselves as playing a zero sum game at this point, and should cooperate more.
By the way, I find it odd that the letter says that Reagan would still have won in 1980 with approval voting. I find this prospect especially dubious if more candidates had run, which would have been likely under an approval system. That statement seems slightly at odds with the authors' claim that approval helps to promote the election of centrists. --Hermitage 08:27, 15 July 2005 (UTC)
Yeah, that is interesting; I'm not sure what the point was there. Anyways, I would like to cooperate with supporters of IRV because they have built up incredible momentum. However, there are so many flaws with IRV that seem to be easily fixed under a Condorcet method (I usually suggest Schulze method). The usual criticisms of Condorcet ring hollow to me: that it doesn't express "strength of support", but only breadth of support - no, the method just doesn't discard all the other preferences you've voted for beyond the first or second place candidates. There are so many examples of where the result of IRV makes absolutely no sense - I like the second example here, where if voters vote twice according to their sincere preferences, they vote the same candidate both as the worst and the best candidate. I just haven't seen any major practical flaws of Condorcet like that. However, I feel that this all might be in vain if the nation refuses to support a ranked choice ballot. That's why I will usually only mention my support for approval voting unless the person I'm discussing with has already heard of alternate methods. Approval voting is a simple change ("choose one" - "choose one or more"), and it yields plenty of benefits. I don't believe there's any good reason to choose plurality over approval voting, which is why I'm surprised AV movements haven't gained more traction. I guess it's just a matter of the motivation of the supporters - IRV seems to have a lot of wind in its sails. - McCart42 (talk) 15:22, 15 July 2005 (UTC)
I hope that nobody minds us using this board to go a bit off-topic. I'm happy to talk about this, though, because I think that this is important.
First, IRV surely has its limitations, but I think that it is quite clearly better than plurality and runoffs. It probably won't really get rid of the two party system, but it should at least give third parties some kind of a toehold that they don't have currently. Some people argue that IRV will get in the way of more effective reform. Personally, I doubt that. I find it more likely that it will engage people's interest in alternative voting systems, and bring the issue into the mainstream.
As for approval, I agree that it's clearly better than plurality as well, and I also find it surprising that it hasn't made much progress politically. Hopefully that will start happen in the near future. Of course, like IRV, approval also has its limitations. I imagine a fairly large amount of strategic guessing and second-guessing, that may work out well at some times and not so well at other times; that is, there may be many cases where voters will regret either approving or not approving a given candidate after the election. Some people think that approval voting might be struck down by the courts on the grounds of the "one person one vote" interpretation of the constitution. IMO, an enlightened judge should realize that approval is constitutional, but we can't necessarily count on that.
As for Condorcet, Smith efficiency is one of my primary long-term goals for public election methods. I think that the election of majority-supported non-polar centrists is badly needed at this point in history. IMO, Smith efficiency is the best definition of majority rule. The main problem is that Condorcet methods are vulnerable to the burying strategy, by which a very clear majority beat can be overturned. See for example: 1, 2, 3, 4. IMO, some Condorcet methods are less vulnerable than others. WV is less vulnerable than margins, but only if people do a fairly large amount of truncation. IMO, cardinal pairwise and approval-weighted pairwise reduce the vulnerability to an acceptable level. (Maybe I'm biased, because I invented them.) There is a difficulty in that the less vulnerable methods tend to be more complex to explain, which may pose a significant political barrier in the short term.
So, I'd ideally like to start by implementing both IRV and approval voting in various locations around the country, and implementing a non-binding direct democracy system that uses a pairwise tally. Then, in the long run, it might be possible to begin using some of the better Condorcet methods to elect executives. --Hermitage 02:34, 16 July 2005 (UTC)

These are very good points, and ones that I hadn't heard before. I haven't read examples 2, 3, or 4 yet, I will do that soon. I think that in terms of changing our deeply entrenched methods of voting on a national level, we might need to make baby steps. In this respect, the "best" voting method might be too much of a leap for people to understand if it is too complicated. Until now, my opinion of the "best" method was a well-designed Condorcet method; now, I have my doubts and will need to do a great deal more reading. That's the point, though; if a college-educated person has to sit down and read several pages of examples, think about it for several days, and discuss it with others to get to the truth of the matter, does it have any hope for adoption on a national level, in an arena that is currently dominated by talking heads that really like to use scare tactics and 5 second sound bites to get through to their audiences? I am not at all trying to attack the intelligence of the average American. What I'm trying to say is that it might be an uphill battle in any larger arena than a citywide election system. So, perhaps looking at ease of transition is just as important as looking at mathematical definitions of fairness. I don't tend to compare methods against each other as much as I do against the current system. If we have to move from approval to IRV to Condorcet, so be it. For all the weaknesses approval voting has, it is pretty much no worse than plurality in any respect except the majority criterion...which, I think, is OK if voters understand that approving a candidate means that they would be happy if that candidate won, period. And approval voting would be an almost trivial changeover from the current plurality system - the only major issue to address is public perception. - McCart42 (talk) 04:44, 16 July 2005 (UTC)

After reading through the example, I've decided that burying doesn't worry me as much as nonmonotonicity - with caveats, which I will mention towards the end of this response. The difference is that in the popular burying example, by strategically voting B>C>A rather than B>A>C, you are saying that you would rather take the chance that C would win in order to be sure that A doesn't win. You're not so much gaming the system as expressing a different preference; this could backfire on you if it isn't your real preference.

Let's look at the result of changing your preference: you give A a lesser chance of winning, and increase C's chances of winning. The surprising side effect is that by not changing B's chance of winning, you've managed to put A below B. Personally I'm somewhat able to reconcile this. This burying, or violation of later-no-harm, as I understand it, is incompatible with the Condorcet criterion -- if I had to pick one of the two (which apparently I do), I would choose to comply with Condorcet.

I suppose there is the risk/reward balance to consider as well, in that burying isn't a large risk while toying with monotonicity under IRV is a large risk. It certainly demands some more thought. I would like to put later-no-harm criterion up as an article so we can discuss it some more; anyone already tried this? - McCart42 (talk) 05:59, 16 July 2005 (UTC)

I think that perhaps invulnerability to burying should be a criterion by itself; I don't think that later-no-help and later-no-harm say exactly the same thing. I like the secret preferences criterion, which I list on my site as the [2] later preferences criterion.
I agree that risk/reward ratio is a key thing to look at when assessing strategic behavior. I think that most people regard the "push-over" strategy in IRV as usually having a higher risk/reward ratio than the "burying" strategy in Condorcet, and as being more difficult to coordinate. The cardinal pairwise method attempts to make it so that few potential strategizers see a low risk/reward ratio, and so that the most flagrant strategic incursions are well guarded-against.
I've never really decided whether the burying vulnerability in winning votes Condorcet is acceptable. (I feel fairly sure that it is unnacceptable in margins Condorcet.) It can be argued that it is fairly burying-resistant in cases with only two major candidates, under the assumption that voters consistently use a truncation strategy. This means that it may work "well" in as least as wide a range of cases as IRV...
Honestly, voter strategy is a truly, wonderfully baffling issue. Perhaps one cannot gain a genuine understanding of strategy in a given method before that method is put into practice in high-stakes elections. I don't claim to know how likely a successful burying strategy will be in a large election, but I do think that there could be a lot of anger if one succeeded in switching the result to the polar opposite of the sincere winner. Hence I can't really offer a firm conclusion about the viability of WV methods at this time. At this point, I would rather see them for shorter-term offices such as US Congress, rather than longer-term offices such as US President or US Senator. --Hermitage 10:00, 16 July 2005 (UTC)
The one thing that somewhat concerns me about the main burying example is that A votes over B only has a margin of 2 votes; it's a very tight race to begin with; thus, when every single B over A voter changes their strategy to burying, B wins. I don't know, it might be the case that the simplicity is purely for demonstration purposes, just as many of the Condorcet over IRV examples are, but it seems like a very close race to use as an example. I wouldn't call it a travesty if B won in that tight a race. That'd be like saying Gore defeating Bush in the 2000 election would have been a complete misrepresentation of voter preferences. It could've gone either way. Maybe with a better example, the burying problem will become more evident to me. - McCart42 (talk) 22:27, 16 July 2005 (UTC)
Additionally, which Condorcet methods rely on winning votes to discard ambiguities, and which rely on margins? Could we add that bit of differentiation to all the Condorcet methods? - McCart42 (talk) 22:33, 16 July 2005 (UTC)
The Condorcet methods listed on Wikipedia can be used with either WV or Margins, but presently the WV interpretation is the only one covered in those articles. I don't mind this. KVenzke 14:20, July 17, 2005 (UTC)
I've added a brief section on WV and margins to the Condorcet method page. I don't get into the pro and con issues, but at least I define the terms and give an example. --Hermitage


Condorcet loser and IRV

It should be noted here that IRV standardly would fail condorcet loser as it allows a voter-truncated ballot. That is pretty easy to see in that the condorcet loser and be unanimously approved and voted for, whilst the other candidates may or may not be voted for as alternates. Thus, there may not be enough candidates marked on ballots to receive transferred votes from the cl, so the cl can win. The only country that does not allow voter truncated irv is australia. --Fahrenheit451 18:36, 5 September 2005 (UTC)

Condorcet Loser is usually defined on votes as cast, so that if you truncate a preference for someone over the CL, and the CL is elected as a result, this is not usually considered a failure of Condorcet Loser. KVenzke 22:34, September 5, 2005 (UTC)

If they are rated as 0 or not ranked, they are disapproved, and would still figure in the mix. You cannot ignore such candidates. --Fahrenheit451 23:05, 5 September 2005 (UTC)

IP207.59.125.242 changes on 11/5/05

I revert changes, and comment below:

Under:

 Special cases of IRV eliminations
   Consider multicandidate elimination of weak candidates as the first step:

ADDED PARAGRAPH

  • Condorcet method to break tie: Eliminate all other candidates and do a pairwise

comparison between the two. The winner of this head to head race would advance. This is the most democratic method and is recommended by many.

COMMENT: This section doesn't deal with ties. Even if it is moved to the next section (Tie-breaking rules:) Condorcet-based elimination fails to meet a runoff expectation that each voter only has one vote each round.

ADDED JUDGEMENT bold

  • CANDIDATE COUNT: Define a maximum number of candidates that can survive the first round. This method is the least favorable as it limits who can win. If you have many candidates remaining, and a tie between the bottom two, those candidates should only be eliminated if there is no possibility they could win.

COMMENT: This section lists options that are used. The argument of "least favorable" and "should only" are POV that don't belong here.

For these reasons I completed reverted changes by IP207.59.125.242

Tom Ruen 12:41, 5 November 2005 (UTC)