Talk:Approval voting/Archive 1

This is an archive of past discussions about Approval voting. 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.

Archive 1

Archive 2

Archive 3

Archive 4

Condorcet

Hi, I am Mrprasad. I originally added the section "Condorcet Criterion Issue" a few days ago (i.e. 1-31-06; today is 2-3-06). A reader informed me that he was concerned that its use of the term "greatest possible consensus" was POV. I understood that concern and want to edit it in a manner that gets around this issue. However, I have not had the time to search for the necessary sources to back the claims. I do not feel I can rewrite the section without some reference to "greatest possible consensus". Thus, I feel the need to drop the section until I can better source the section. I will return it once I do that.

It is easily reversed as disapproval voting where a choice is disavowed, as is already required in other measures in politics (e.g. representative recall).

I don't see why approval voting is any different than any other system for "disapproval". The case cited (representative recall) is not even a single-winner contest. (throw all the bums out :-). It is a single-candidate contest if anything.... I propose dropping that text. NealMcB 18:10, 2004 Apr 29 (UTC)

Reads well now. The point about approval and preference voting both being single-vote systems is now there, but no hierarchy of 'this is a type of that' is anywhere in the descriptions of the voting systems themselves, which is fine, except for disapproval as a variant of approval, which seems the only clear inverse. 15:02, 2003 Jan 17

There was an earlier article here once, but it must have been censored.

Under Approval_voting#Potential_for_Tactical_voting I don't like this statement:

A good strategy is to vote for every candidate the voter prefers to the leading candidate, and to also vote for the leading candidate if he is preferred to the current second-place candidate. When all voters follow this strategy, the Condorcet winner is almost certain to win.

It is an opinion statement and I don't know what qualifications should be made for it. Who says this is a good strategy? How do we measure the "leading candidate"? Is that an approval measure or a plurality measure?!

I've done my own thoughts on Approval Voting and I judge:

A good strategy is to vote for the same candidate you would vote for in a plurality race, and also any candidates you prefer more (ignoring any assumption that those candidates can't win).

thus reducing the problem to the IMO harder one of deciding whom to vote for in plurality. --Townmouse 19:48, 16 Oct 2004 (UTC)

This stategy is more practical, and is MUCH more likely to be used, even if it doesn't help the Condorcet Winner.

It would merely allow weak candidates a fair measure of support (approval) without throwing the election against a preferred stronger candidate. Approval offers little to voters in a strong three-way race when the top leader is not predictable. Then approval is only useful as a maximal power "negative vote" against feared competitor(s). (Voting A=1,B=1,C=0 in approval is mathematically equivalent to A=0,B=0,C=-1)

I won't change anything for now, but I appreciate opinions. I accept the existing "good strategy" statement as "Good strategy to elect a Condorcet Winner".

Does anyone else have a problem with the existing "good strategy" statement? Any suggestions for changes?

Thanks! --Tomruen 00:02, 2 Jun 2004 (UTC)

The problem with your phrase is that the vote in a tactical vote in a plurality race may also depend on identifying the leading candidates in a plurality race, the leading candidates may be different under different rules and the information may not be avialable on how others might vote in a plurality race. For those who wish to vote in a way which is unlikely to affect the result, tactical voting advice is unnecessary, and I think the current text makes reasonable sense - the leading candidate obviously means in an approval measure. --Henrygb 23:16, 26 Jun 2004 (UTC)

Hmmm... my first assumption was "leading candidate" meant leading in a single-vote count since it is pretty much meaningless to poll voters with approval votes in a poll. Approval polling is too strongly open to manipulation, even unorganized manipulation - individuals playing around for fun or profit. I accept the same argument exists with single-vote counts, but people still want their true favorite to poll well and are less likely to play unless they really have no preference. --Tom Ruen 05:37, Jun 27, 2004 (UTC)

Surely a "good strategy" is a strategy that maximises that particular voter's expected utility. For a single-winner election that means one that maximises how much the voter likes the ultimate winner.

My question is why approval voting is single winner only. Is it called something else when the top n candidates are elected? --Henrygb 23:16, 26 Jun 2004 (UTC)

Approval certainly could be used for multiwinner elections and doesn't have or need any different name. In practice there's less reason to need it. The "standard" multiwinner election method is called Plurality-at-large or Bloc-voting which allows as many votes as candidates. It's less often people would want to "overvote" with multiple winners except as a defensive vote against one or more most disliked candidates. A more likely vote in plurality-at-large is voting for less candidates than winners because you don't want your lower choices to defeat higher ones.

Actually multiwinner Approval elections are an attractive idea - but only when candidates are truly evaluated independently - that is if the number of winners can be variable and the approval rating needed is fixed before the election.

For example a great application would be a poll for which ALL candidates with more than 15% approval are included in a presidential debate. That's a very valid poll since you know the more you vote for the more winners there will be and your vote for a second favorite CAN'T hurt the chances of winning for a first favorite. --Tom Ruen 05:37, Jun 27, 2004 (UTC)

Approval Voting and the Florida 2000 Butterfly Ballot.

Some thoughts:

If Approval Voting had been in force in Florida 2000, then the mistakenly double marked ballots (for Gore and Buchanan) would still have counted. No votes would have been voided. Gore would have got all his genuine votes, while Buchanan was so far behind it would not have mattered it he got some bonus or unintended approval votes.

Does Approval Voting comply with "one man one vote"? Yes it does, because any Buchanan votes do not really matter because he was so far behind, and Gore only ever recieves "one useful vote per voter".

Has Approval Voting ever actually been used anywhere?

The biggest disadvantage of Approval Voting may be that it distorts the election statistics - the total number of votes may appear to add up to more than 100%. Something would need to be done to "normalise" the results.

If the votes in an Approval Vote contest add up to more than 100% turnout, how does convince the public that the election has not be tampered with? That might be a hard sell!

If 999 Nader voters had approved of Gore, Gore would have beaten Bush.
If 999 Buchanan voters had also approved of Bush, Bush would still have won.
Note that there were 3rd Party candidates spoiling things on both sides.

Syd1435 09:23, 2004 Oct 4 (UTC)

The dilemma with Approval Voting

Suppose one side of politics is a permanent Coalition, say City Republicans and Rural Republicans, and an election is being held using Approval Voting.

If you do not mind which kind of Republican gets in, you approve of both kinds.

But say for some reason or another, you would much prefer the Rural Republican to get elected - how then do you fill in your Ballot?

Answer, you must approve of the RR candidate and disapprove of the CR candidate.

Suppose there are other voters who prefer the City Republican candidate.

Taken together, the two Republican candidates will lose a few votes compared to the sole Democratic candidate.

There is thus a dangerous split in the Republican camp, albeit a modest one.

This illustrates a flaw in Approval Voting, because all approvals have the same value, when in practice, some variation in the approval rating is desirable.

What is needed is some kind of intermediate approval level. Here is a sample ballot paper, to enable both Republicans to be approved, but with one preferred to the other. Then all Republican votes count toward beating the common enemy.

[3] Joe Smith ----- Greens.
[1] Henry Ford ---- Rural Republican
[3] Jane Doe ------ Democrat
[3] Fred Rubble --- Reform
[2] Mary Hill ----- City Republican

Notes:

[1] = 1st preference Approval (YES-FULL)
[2] = 2nd preference Approval (YES-PART)
[3] = Disapproval............ (NO)

Note that the Modified Approval Voting with variable levels of preference is starting to look like Instant Runoff Voting.

Put this the other way, Approval Voting is a restictive version of Instant Runoff Voting where instead of having N different ranks for N candidates, you are restricted to 2 different ranks for N candidates. —Preceding unsigned comment added by Syd1435 (talk • contribs) 10:03, 4 October 2004 (UTC)

It would be more accurate to say that Approval Voting is a restrictive version of Range Voting; the two methods are fundamentally similar in process to each other, in a way that IRV isn't. Jack Rudd (talk) 21:59, 15 July 2008 (UTC)

selling IRV

Hi Syd1435,

I'm moving your discussion of Approval voting's flaws to the talk page for two reasons (1) because it really looks like you're actively trying to plug IRV [instead of dispassionately describing the flaws in AppV] (2) the tone of the content is really a lot more conversational than it is encyclopedic. Let's work together to bring the most useful content from your contributions back to the article. Best, -- Kowey 08:55, 16 Oct 2004 (UTC)

P.S. I will confess that I am partial to AppV myself, so we should also get help from a more neutral 3rd party.

Since AV is also an abbreviation for Alternate Vote aka IRV, can we abbreviate Approval Voting as AppV?

I think it helps to compare FPTP, AppV and AltV aka IRV rather than dealt with them dispassionately on their own.

BTW, the 9 Oct Election is very exciting because preferential voting (IRV), slow absentee and postal ballots, mean that unexpected and interesting results are occurring. No point in having it all over on election night.

I am happy to work with you to get something acceptable back of article page.

Syd1435 12:10, 2004 Oct 16 (UTC)

One Man One Vote

In an Approval Vote election, the total number of approvals will add up to more than 100% of the number of voters, which may appear to some people as a breech of the "One Man One Vote" principle. It is not a breech, but how does one argue that it is not a breech?

(note: using a different colour to set discussion apart from content)

One man one vote means that nobody has more voting rights than anybody else. Seems perfectly intact: everybody has the same ability to choose as many candidates as he wants. Another way to look at this issue is to think of "not voting for" a candidate as equivalent to "voting against" that candidate. In other words, everybody gets one vote on each candidate: yes or no. In any case, this needs to be massaged into encyclopedic form or left out -- Kowey 09:26, 16 Oct 2004 (UTC)

I agree that FPTP, AppV and IRV are all "One man one vote meaning that nobody has more voting rights than anybody else." However, during the Alaskan referendum for IRV, the AK League of Women Voters (though not the national LWV) questioned whether IRV was compatible with one man one vote. They got no satisfactory answer, and so opposed any change. This was very disappointing. The AK LWV thought that if your preferences were transferred N times, that was one man N votes for some, while only 1 man 1 vote for others. A better answer to the AK LWV might be to say that no matter how many times preferences are transferred, the total number of votes remains constant, thus one man one vote.

Syd1435 10:45, 2004 Oct 16 (UTC)

Suppose that a voter approves of candidates B and C; but approves of candidate B more than C. How does he cast his votes?
- If he approves of only B, the lack of an approval vote for C may let A in.
- If he approves of both B and C, then he has failed to express his preference for B over C.
- It is thus impossible to simultaneously approve of two candidates which preferring one over the other.
- Approval voting suffers the disadvantage that all approvals are of equal weight, even when the vote wishes to express unequal approvals.

Tactical voting

I think it is a mistaken assumption that accounting for (proportional) preferences between candidates is even neccesary or desirable, but that's just my opinion. Seems like the most important thing in an election is to prevent tactical voting, and allow voters to vote sincerely for who they want without having to perform any calculations. That being said, I think the article already mentions that AppV does not account for preferences, but IRV does, but then again, IRV has its fair share of problems, in that attempting to account for preferences, it also introduces some potential for ~~strategic~~ tactical voting, no? -- Kowey 09:26, 16 Oct 2004 (UTC)

I assume that tactical and strategic voting are more or less the same thing, choosing a lesser choice, to avoid a worst choice. IRV is the one system that avoids the need for any of this negative tactical voting. You rank all the candidates, and the full value of that vote stays still or shifts until either it is victorious or is defeated.

As is being shown in the 9 Oct 2004 election, if there are a lot of candidates competing for a seat, the flow of preferences are too complicated for mere onlookers (and even the candidates themselves) to really know what is going on, but all will be revealled as the Electoral Commission crunches its numbers.

PS how do I change the background colours?

Syd1435 12:24, 2004 Oct 16 (UTC)

I am using tactical/strategic voting interchangably. If there is any difference, somebody shout. For now, i'll try to use "tactical voting" strategically.

Anyway, I'm not trying to dodge the discussion here, but we need to keep in mind that the goal is not to determine which is the superior voting system, but to present the user with descriptions of them, as well as a fair discussion on their relative strengths/weaknesses. Yes it's tough to draw the line between such a discussion and the one we are having now, but we should be making a stronger effort to avoid steering the user to any conclusions which are based on opinion or ideals.

Also, No need to change background colours and/or create new div tags. My only goal was to set our discussion aside from your original text.

-- Kowey 13:13, 16 Oct 2004 (UTC)

These two problems are resolved with Instant Runoff Voting.
- With IRV, the total number of votes for each candidate at every stage of the count equals the number of votes orignally cast.
- If a votes put B first, C second and A third, then the full value of his vote is set against B in the first round. Depending on how B and C score against eachother, either the second preferences of C go to B, or the second preferences of B go to C. No votes need be lost in the struggle between like minded B and C against the common enemy A.
- In military terms, there is no "Division of Force" when a voter preferences B before C. No matter whether a votes remains with its original choice, or is redistributed to a lower choice, that vote always retains its full value, one man, one vote, one value.

There is a problem with Instant Runoff Voting that in practice is no longer a problem.
- With Approval Voting, Limited Voting, and the like, when a computer is used to tally the results, only the total score for each candidate need be recorded. This requires only a small Commodure 64 computer with 64k of memory.
- With Instant Runoff Voting, it is necessary to store an image of every ballot; it there are N ballots, and M candidates per ballot, then a Pentium computer with N * M bytes or even double bytes of memory is required.
- It can also be represented with much less memory. Just as the example above represented many voters with just four columns, you can get my with just counting how many ef each *unique* ballot exists. You only need M * (M!) entries, each entry able to count up to N. For M=4 candidates, this comes to 96 entries. For M=10 candidates, this comes to about 36 mega-entries.
- If Department of Elections only has Commodore 64 computers at its disposal, then a computerised IRV may be out of the question.
- A computerised count of and IRV election would also take much more time to process than an Approval Voting election.
- How many people still use Commodore 64 computers?

I realise you're just trying to make a concrete example here, but I think talking about Commodores and Pentiums is a bit completely off topic, and for that matter, underestimating the power of the Commodore 64 (and or paper/pencil). -- Kowey 09:26, 16 Oct 2004 (UTC)

Put simply, IRV requires several orders of magnitude greater computer power to process an election than FPTP, AppV, CumV, LimV. These less demanding systems only have to process each ballot once, can total things as they go, do not need to record a copy of any ballot (except for auditing purposes), and the software would be small and simple.

IRV is the opposite. However IRV computers do exist, and large enough computers also exist. IRV can still be done on paper ballots.

Syd1435 10:56, 2004 Oct 16 (UTC)

I think your orders of magnitude come from a poor choice of algorithm. The number of ballots in IRV is irrelvant; you only need to store M^2 numbers: for each candidate, the number of 1st rank votes, 2nd rank votes, etc. In other words, IRV is simpler to implement than you suggest, so be happy, but all this talk of complexity is not relevant to article on voting systems. -- Kowey 11:20, 16 Oct 2004 (UTC)

I am trying to concede that IRV requires more resources than other systems, and if these resources are not available, then a system like AppV would be a satisfactory second preference. If you are not bothered, fine. :-)

BTW, information must be stored for each of the N ballots, because with IRV, unlike FPTP or AppV, you cannot just add up the numbers on the ballot papers as they go by. These ballots must be reprocessed as many as M-1 times to attain a result.

Fine, my mistake; i hadn't understood how the IRV transfer worked and after rereading carefully, I will tentatively agree that you will need to store each ballot. But no, I am not bothered, and concession is not neccesary because I think it's not really pertinent to the article. -- Kowey 13:13, 16 Oct 2004 (UTC)

I would like to point out that IRV can also be exploited through tactical voting whenever the Voting paradox applies. No ranked method is immune to the Voting paradox. Paladinwannabe2 19:56, 11 October 2007 (UTC)

Disadvanges with CAV sample ballot

The Citizens for Approval Voting website uses the following ballot paper as its logo:

- [YES] Jefferson - a great president
- [NO ] Stalin - a tyrant
- [YES] Madison - a great president
- [YES] Washington - a great president

Clearly, it does not matter which of these three great presidents wins, so long as Stalin loses.

But suppose the ballot was revised as follows:

- [YES] Jefferson - a great president
- [NO ] Stalin - a tyrant
- [YES] Madison - a great president
- [ ? ] Warren G. Harding - a ROTTEN president

Now the choice is less clear. We would want even Warren G. Harding to beat Stalin, but either Jefferson or Madison to beat Harding. This example shows a fundimental weakness of Approval Voting. Cumulative Voting does better, and Instant Runoff Voting better still.

Uh no? If fewer people like Harding, fewer people vote for him. Jefferson/Madison still win. Problem solved? -- Kowey 09:26, 16 Oct 2004 (UTC)

Kowey overlooks the point that only Harding is contempory with Stalin, the other two presidents being dead. And what if the choice is between the worst three presidential candidates, say Harding, Dukakis and say Hoover?? -- Syd1435 20:26, 16 Oct 2004 (AEST)

- [ ? ] Dukakis - an unsuccessful presidential candidate
- [NO ] Stalin - a tyrant - was very "good" at being a tyrant.
- [YES] Hoover - a unsuccessful president ?? excellent mining engineer in Australia and elsewhere.
- [ ? ] Warren G. Harding - a ROTTEN president

Ok, here i'm confused: what does Harding's being contempory with Stalin have to do with anything? We're talking voting systems here with a hypothetical example, and for some reason or another the voter is presented with 4 choices. Easy enough, the voter selects his two acceptable choices. I also do not see how the case of four undesirable candidates (of which one or two really not desirable) is any different. You just pick the lesser evil(s)... or you file a protest vote by voting blank. -- Kowey 10:51, 16 Oct 2004 (UTC)

Some additional points:

Approval voting is better than First Past the Post. FPTP is like a car with no gears. No Gears = Very Low Performance.
Approval voting is simple and inexpensive to implement, like a car with manual gears. Some performance.
Instant Runoff Voting is complicated and requires more resources, like a car with automatic gears. High Performance.
But IRV has been shown to work, in the same way that automatic gears for cars have been shown to work, and is worth doing.

-- Syd1435 20:26, 16 Oct 2004 (AEST)

Sure, I'll agree that AV and IRV are preferable to FPTP. But I suspect that you are over-estimating the gains from IRV (i.e. ability to rank your preferences) because you are not considering that these gains are offset by increasing the likelihood of strategic voting. In other words, AV's simplicity is an inherent virtue, not just an implementation advantage; it is, in a sense, more fool-resistant. But this is an opinion only; i don't have the real world examples to back this up. The CAV have a page that mentions this re-emergence of Wasted Vote Syndrome. They claim that this leads to strategic voting in practice (i.e. in Australia), exactly what I want to avoid.

That being said, the point here is not to convince the reader that we should use one system or another. We are an encyclopedia. The task at hand is help the user learn about the systems, how they work and what the basic implications are; it is not to sell them anything. -- Kowey 10:51, 16 Oct 2004 (UTC)

Put simply, Approval Voting cannot simultaneously approve of multiple candidates and choose between them.

An even more telling ballot would be to place Stalin against the three worst presidents.

Strategic or Tactical Voting

Strategic or Tactical Voting occurs when a voter judges that their preferred candidate is not likely to win, and that therefore they are better off voting for a lesser of two evils candidate.

With FPTP, this can occur with three cornered contests. An unexpected third candiate can drain away enough votes to spoil victory for one of the leaders.

With Approval Voting, the drainage of votes is certain to be much less, since a voter can select both the original choice, as well as the lesser of two evils choice.

If however, the vote judges than their preferred candidate has a winning chance, they may decline to approve of the lesser of two evils choice, making room for the worst of two evils to get in, if they miscalculate. This could be gut-wrenching. Some voters will thus only approve of the preferred candidate and decline to support the lesser of two evils candidate, casting a FPTP like ballot.

With Instant Runoff Voting, the voter ranks their first choice first, the lesser of two evils candidate second, and the worst of all evils last. No combination or permutation of preferences can help elect the worst of all evils, assuming that the voter ranks all candidates. The need not be any leakage of preference that might spoil the result as a PFTP vote can do, or a AppV vote does to some extend.

When an IRV ballot is filled in, with all the candidates ranked consequtively from 1 to N, there can be no circular preferences, (A>B, B>C, C>A). Preferences always flow unambiguously D>A>E>B>C, etc.

With AppV ballot there are no circular rankings, but there are ambiguous rankings say (A=D=E)>(B=C).

With FPTP ballot, there are no circular rankings because say D>(A=B=C=E).

What is this ranking circularity called - Concordat blah blah???

Syd1435 11:48, 2004 Oct 16 (UTC)

Syd1435, please read the writeup on Instant Runoff Voting as well as the writeup on Condorcet method for an explanation of IRV's flaws, and a demonstration of why IRV has major tactical voting problems when a third party candidate reaches a level of popularity where it is possible that they could be elected. The point to take home here is that approval voting isn't perfect, but it is better than first past the post, and a good first step towards a true Condorcet method for elections. Instead of Instant Runoff Voting, perhaps you would approve of a ranked choice ballot, counted with Condorcet method? McCart42 20:49, 2005 Mar 9 (UTC)

Abd 15:57, 20 December 2006 (UTC): I deleted the following section:

The mathematics of approval voting lend it to some manipulation and tactical voting. As each vote counts as one vote and the winner is the one with the highest total, each vote equally helps the candidate/issue (city in this example) selected win. Because of this, voters are more likely to only vote for their favorite. Because Approval voting has not been used much for real elections, this phenomenon is not well documented.

The reason is that it is redundant, where it is accurate, and POV where it is not. Specifically, the paragraph claims that, with Approval, voters are more likely to only vote for their favorite. (1) this is speculative and clearly POV. (2) It is highly unlikely to be true. Some voters will do this, for sure, but others, and especially the supporters of third party candidates, will follow what has been recommended as a standard Approval strategy: vote for your favorite(s), vote for your favorites among the leading contenders whom those you think have a shot at winning, and vote for any candidate you prefer to a favorite leading candidate.

Further, many voters who have a favorite who is a leading contender may well also vote for a candidate whom they wish to encourage in some way. So, for example, a Democrat might also cast a vote for a Green candidate in order to indicate support for more liberal policies, thus possibly encouraging the Democratic Party to move in that direction.

In addition, the statement uses "more" without any reference to what is being compared to Approval. Abd 15:57, 20 December 2006 (UTC)

Like any Preferential voting system, IRV can be exploited through tactical voting. given the classic example of

7 voters prefer A > B > C

6 voters prefer B > C > A

5 voters prefer C > A > B

A would normally win in an Instant Runoff election. However, if 3 voters who preferred B marked down (C>B>A) on their ballot, then C would win the election instead- so by changing their vote, B's supporters have kept their worst-case candidate form being elected. Paladinwannabe2 20:15, 11 October 2007 (UTC)

Instant Runoff Equivalent of Approval Voting

Approval Voting can be converted into a form of Instant Runoff Voting by replacing YES with "1" and NO with "2"

[2] Joe Smith ----- Greens.
[1] Henry Ford ---- Rural Republican
[2] Jane Doe ------ Democrat
[2] Fred Rubble --- Reform
[1] Mary Hill ----- City Republican

Note how the numbers "1" and "2" are repeated.

If a voter desires to preference Ford ahead of Hill, while supporting both these Republicans against all others, then that voter would need to use the numbers "1", "2" and "3".

This "1" = "FULL YES", "2" = "PART YES", "3" = "NO"

[3] Joe Smith ----- Greens.
[1] Henry Ford ---- Rural Republican
[3] Jane Doe ------ Democrat
[3] Fred Rubble --- Reform
[2] Mary Hill ----- City Republican

It would not be Rocket Science to devise a (computerised) system to allow a voter to use either Approval Voting or Instant Runoff Voting in the same election.

[3] Joe Smith ----- Greens.
[1] Henry Ford ---- Rural Republican
[4] Jane Doe ------ Democrat
[5] Fred Rubble --- Reform
[2] Mary Hill ----- City Republican

A convential Instant Runoff Voting requires each square to be filled in by different consequtive numbers, which helps the electoral officials check for errors.

I don't see how this is relevant to the article, or to one about IRV. -- Kowey 09:36, 16 Oct 2004 (UTC)

Multiple winners

Approval voting can be extended to multiple winner elections, either as block approval voting, a simple variant on block voting where each voter can select an unlimited number of candidates and the candidates with the most approval votes win, or as proportional approval voting which seeks to maximise the overall satisfaction with the final result using approval voting.

A disadvantage of multiple winner voting, is that if there are N vacancies, voters must vote for N candidates, which is like N First Past the Post votes in parallel. The result if often a clean sweep by one major party or the other.

Limited Voting allows each voter M votes for N vacancies, where M < N, which reduces the likely of a clean sweep.

Cumulative Voting allows each voter M votes for N vacancies, where M <= N, but the voter can allocate those votes 1 each for M candidates or M votes for just 1 candidate, or something in between.

AV is an abbreviation of "Alternate Vote aka IRV" as well as Approval Voting. Can something be done to disambiguate this, by say calling Approval Voting AppV?

Syd1435 10:56, 2004 Oct 16 (UTC)

The link to "The Science of Elections" requires a subscription. Is there a summary anywhere that can be viewed freely? --Townmouse 19:48, 16 Oct 2004 (UTC)

Quibble

In contentious elections with a super-majority of voters who prefer their favorite candidate vastly over all others

Why do you want to use the term "super-majority" here? The old text said "full of." If you don't like that, then how about "many"? KVenzke 19:01, May 22, 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) 04:11, 15 July 2005 (UTC)

Elaboration of Tenessee example

I'm not too happy with how the Tenessee example is elaborated in the article page, in particular the sentence that ticks off the elaboration of that example:

Supposing that voters voted for their two favorite candidates, [...]

While that's typically not the kind of simplification approval voting makes.

I can understand a more complex example (than the new "simple example" I introduced) is needed to clarify how the "tactical" voting works, and other concepts related to approval voting that need explaining, I can even appreciate such "untypically simplified complex examples" come from handbooks written by serious people, but in the end I think it might be perfectly possible to lend & adapt an example (or set up one) that neither has the flaw of not being a real approval voting example, neither is too limited to explain the most important concepts relating to this type of voting.

For clarity: I have no problem whatsoever with Tennessee or whatever other state or country being used by way of example: only the example should not set the bad example (or even merely suggesting) of making the supposition that in an approval vote everyone votes for exactly two candidates.

I'd rewrite the example, but then I have no clue what even remotely likely data might have been for Tenessee; and also spilling these data in the article page might encumber the example with a degree of complexity not useful in explaining the base concepts.

Could anyone help out here?

--Francis Schonken 10:44, 25 August 2005 (UTC)

I don't agree with you that the Tennessee example is misleading. The only thing that could reasonably be done would be to consider also the cases that voters approve the top one or three candidates. A mixture of strategies wouldn't explain anything. KVenzke 04:34, September 12, 2005 (UTC)

"Live voting"

User:Couvares added a bullet point: Approval voting is simple enough to carry out quickly in a live meeting without the aid of computers, complicated paperwork, or multiple rounds of voting. This makes it more practical than many other voting systems for small, democratic organizations such as labor unions.

I removed this opinion statement of "more practical" which ignores the difficulty for voters in quickly deciding how many candidates to support.

I'd actually be interested in separating two methods that fit under "approval":

Approval voting with a fixed number of winners.
Approval voting with fixed threshold for acceptance of a variable number of winners.

They are categorically different in voter strategy.

I consider Fixed-winner Approval Voting in live meetings a perfect example of how not to use approval.

I've seen it attempted. A dozen candidates were sequentially listed with people raising hands on each candidate for support. Voters couldn't easily keep track of the full list of candidates in their heads without anything written down. I believe candidates named near the end of the voting got less votes as people got saturated by choices. I believe other more careful voters (like myself) offered to few votes I was I unable to backup the voting and add my support for a previously announced candidate. It was a real mess.

On the other hand, a fixed-threshold approval vote might be used to select sequential discussion items:

For example, a progressive/efficient meeting chair might say "Brainstorming has produced 8 independent proposed soluions to our problem. To speed up discussion we'll in immediate sequence yes or no whether further discussion will be allowed for each proposal. Choices with 50% support will have 30 minutes of discussion following. Choices below 50% will be disregarded immediately."

Under a fixed-threshold system, undervoting on "fair choices" to help a favorite choice is discouraged for fear of losing ALL fair choices. Overvoting is discouraged only by the limits of patience of the voters to deal with more winners.

I could see something like this would seem "more democratic" than using simple "Robert's rules of order", with a single "motion" up at a time, and people being forced to vote on a final decision before knowing all the options.

I accept when the term "Approval Voting" is used, it is assumed to mean fixed-winner Approval.

I don't believe there is a practical use for "Approval voting" in live meetings because of the difficulties I offered above. A voting method shouldn't give more power to smarter people - people who can quickly apply strategy on how widely to vote.

Tom Ruen 04:45, 24 September 2005 (UTC)
UPDATED: Tom Ruen 07:12, 24 September 2005 (UTC)

Approval Voting in the real world

The question was asked, where do they use approval voting?

The United Nations uses Approval Voting to select its Secretary-General; this highlights AV's ability to select the least objectionable candidate when voters vote their opinion (as opposed to "strategically voting").

Also, around 500 B.C., Athenians used annual disapproval voting to determine which politician they most wanted to exile (to prevent too many leaders from being banished, the rules were changed to require 6,000 such votes).

From: Fred Baldwin Ph.D, "Election Dissection - Different polls for different goals." July 2004. Attache http://www.attachemag.com/archives/07-04/features/story2.htm

Iarex 22:26, 30 November 2005 (UTC)iarex

Apparently the UN does NOT use multiple rounds of voting (from the link). I wonder if they care what "approval" level the winner gets? Do they care if one candidate gets 90% and another gets 89%? Do they care if no candidate gets 50%? Do they care if no candidate gets even 10%?

I still don't see any "live meetings" successfully using approval.

Tom Ruen 05:24, 24 September 2005 (UTC)

Tom, you seem to have a POV about whether Approval should be used in live meetings, which is not especially relevant to the article. I also see no reason to remove the fact that Approval is easy for live meetings; you can have the opinion that it's flawed, but that doesn't make it less easy.

Approval voting, as defined, has one round; if you have multiple rounds, then you're holding a runoff election that has much different properties. I have only ever witnessed one Approval runoff, and it was here on Wikipedia for the template standardization contest. (It went very badly.)

RSpeer 15:37, 24 September 2005 (UTC)

Given the fact that I have no evidence of Approval being used successfully in live meetings, my opinion would seem to be confirmed that Approval is not easy.

Tom Ruen 18:50, 24 September 2005 (UTC)

Singular they

I had no problem with the voting procedure being described before using the singular they. The "he or she" strikes me as just a bit more awkward, but I don't think it's worth reverting over. rspeer / ɹəədsɹ 19:05, 24 December 2005 (UTC)

Links

I removed the following "studies":

Thomas Colignatus (afaik his real name is Thomas Cool but he prefers Colignatus because he thinks it's cool) PDF: "The normal state in the world is “strategic voting” (cheating) and then AV collapses to “Plurality Voting” (PV), where people basically vote for their first candidate."

Emile C. J. Sheng PDF: "From a game theoretic standpoint, voters with ordinal preferences among the candidates are decreasing the influence of their ballots if they vote for multiple candidates; their votes are most decisive and influential when they concentrate their support on one candidate only."

Incorrect. Example: There are three candidates, A, B, C. A and B are well established. C is a complete newbie with almost no funding. I like C most, B almost as much and I despise A. If I am too proud to compromise so that I only vote for C, I give away the possibility to influence who wins. I will mark B. But making also a mark for C comes at 0 cost, so I will mark B AND C. --R.H. 01:22, 29 January 2006 (UTC)

Those conclusions seem fairly dubious to me, yes, but the fact that you disagree with the studies is not a reason to remove them. Do you have a better reason? For example, have these papers not been peer reviewed and published? rspeer / ɹəədsɹ 07:44, 29 January 2006 (UTC)

Thomas Colignatus says that the status quo should have privileged status in voting systems. In general, giving no option a privilege is viewed as a feature (Neutrality Criterion) and not a bug . One can alter any method to have a status quo bias though.

See also my comment below. On this point here: there is a difference between technical and moral. Technically it is possible not to eat, but that does not mean that you should not eat. Technically you can present voting systems without the status quo given special position, and then it helps to see the features that arise; but morally you would want the pre-condition of the selection of points that are at least as good as the status quo, in order to protect minority rights. Thus, the term 'neutrality criterion' would be a technical phrase that can be misleading about its moral impact. Indeed, you can add tires to any car, but, it would be strange when car sales persons start selling cars without tires, since you can always add them yourself. Thus, voting theorists should think twice before presenting voting systems to the general public without the status quo pre-condition. Thus, I have always granted that 'technical neutrality' is a feature and that the status quo can be added, but my point was that from the angle of democratic theory, the issue is reversed. Thanks for referring to my work, but please refer accurately. Colignatus 22:59, 10 March 2006 (UTC)

Emile C. J. Sheng cites an example of voter preferences (I mean what is in their heads before they vote) outlined by Approval supporter Lin Jih-wen. It is the first scenario in the article. He claims Lin Jih-wen used this as an example how strategy-proof Approval is and then tries to use the same example to show the opposite. The scenario shows a Condorcet Cycle (he doesn't mention that it is one). What method doesn't look bad with a genuine Condorcet Cycle? (IRV for example fails Reversal Symmetry in such situations). Given that genuine Condorcet Cycles happen rather seldom (according to simulations posted on the electorama mailing list those occur with non-strategizing voters with a likeliness of 5%-10%) it probably doesn't tell much positive or negative about Approval.

He proceeds to describe how in that situation voters have to vote somebody over their true favorite to prevent the one they dislike most from winning. "How is that possible?" you might ask, given that according to Mike Ossipoff and others there is no reason to put somebody above your favourite (Favorite Betrayal Criterion). Well, the Condorcet Cycle is not enough for that, you must also change the votes from being cast at the same time with protecting privacy.

I don't think Brams advocates that.

He continues by saying Approval fails the "monotonicity condition". The wikipedia article says it passes. Douglas Woodall has several criteria that he calls mono-something and he calls what is normally called monotonicity "mono-raise". What Sheng means by "monotonicity condition" he never says. --R.H. 19:09, 29 January 2006 (UTC)

Approval Voting & Minorities

I wonder whether non-preferential Approval voting (cf. Preferential voting) has to be (game theory) or usually is against the representation of minorities. Taking an example from e.g. staff representations or union elections: 3000 possible votes, 11 choices on ballot, 7 of those 11 will get a seat, a minority of 500 (17%) which is 100% for representation of 1 person, usually get's not represented.

My question: is this mathematically inevitable?

Is it worse the more unpreferred choices there are on the ballot paper?
If this problem is known, how to improve it?
How to improve it in an acceptable way for the majority - otherwise you don't get it done.

If this problem is already addressed here somewhere, please point to it.

Are the voting simulating programs, where one could test things and experiment with voting behaviour of groups of voters? Tommie 12:26, 30 January 2006 (UTC)

Approval Voting in general means the single winner method described in the article. Multiwinner doesn't necessarily mean proportional.

The first winner with Block voting is probably a great centrist (especially if you are allowed to make as many marks as you want), the second winner makes an almost as good centrist, the third winner is almost as good at being a centrist as the second etc. The reason why simply taking the top vote getters as winners is unlikely to produce proportional results is because Approval Voting is a good method to elect a centrist. This also tells us something about single mark ballots. The reason why lone mark plurality ballots can be used for more proportional representation or runoffs with exciting debates among interestingly different candidates is because lone mark is so bad at electing centrists.

Here is a worst case scenario with block voting: Suppose there is a strong divide with people divided into camps so that everybody in a camp votes for every candidate from that same camp and nobody else. Group A is 20% of the voters, group B 19%, group C 18%, group D 17%, group E 16%, group F 10%. 10 seats to get. If 2 candidatas from group A run, they both get a seat which is fine. If 3 candidates from group A run, they all get a seat. Well, it is very seldom possible to give seats exactly according to the votes, there are basically always rounding errors. But what if 4 candidates from that group run, or 5, or even 10? They will all get a seat.

Marking a candidate that doesn't get a seat shouldn't cost the voter something (like it does with cumulative voting). But a ballot that is successful at electing a candidate for the first seat must be downgraded in voting power for the next round of counting so that voters who got shafted in the first round have more say in the second round which picks the second winner. This is what the method by Thorvald Thiele does. It is described at the end of the Proportional approval voting article (the nonsequential version described before that gives more proportional results but needs a computer). The method by Thiele has been criticized because the amount of downgrading a ballot gets is only determined by how often it is successful at electing a candidate. Suppose there is a highly divided electorate and the first seat goes to somebody only approved on 10% of the ballots. Each of those ballots get downgraded to one half in the second round. Compare that to an electorate that strongly supports one candidate so the first seat goes to somebody approved on 90% of the ballots. According to Thiele each of those ballots get downgraded to one half for the second round as well. E. Phragmén developed a counting method that recognizes the difference: Olli Salmi on (Sequential) PAV and Phragmén --R.H. 15:16, 30 January 2006 (UTC)

I just entered a User talk page http://wiki.riteme.site/wiki/User_talk:Colignatus and, to check that it had been entered correctly, I did a search and also noted this page. Two Points: (1) It appears that there had been a link to a paper of mine, that R.H. however removed again. It appears that I have been quoted correctly that AV "basically" comes down to PV. But R.H. neglects this "basically" and gives an example where there is a small difference. My point however was that, granted that such exceptions exist, the rule is that AV ~ PV. I emphasize this point since my research shows that a Borda Fixed Point approach would be best, considering common conditions, so that all research into AV most likely is a dead end. (2) One question was whether my paper on AV has been peer-reviewed. No, my papers generally aren't, since I wait till the censorship of science by the Dutch government has been solved, and then I can start joining the usual scientific process again. In the mean time I put my work on the web so that people can see what the censorship is about. (1) + (2) I would enjoy if the reference is restored, though will not do so myself, not only for the above, but in particular since the issue of the Status Quo is important in that paper as will, which issues tends to be overlooked in voting theory ("tends" is not "is"). Thanks for the good work. Colignatus 21:38, 27 February 2006 (UTC)

Approval-Disapproval Voting

I've come across an interesting discussion of the possibility of an approval-disapproval voting system. Take this example:

[DISCLAIMER: I'm not a Libertarian or pushing an agenda.]

I'm a Libertarian voter - for both free market and what I view as civil rights. My choices are as follows:

Libertarian Party - Share nearly all values.
Democratic Party - Share some social and economic values.
Republican Party - Share few social values and many economic values.
Green Party - Share many social values and few economic values.
Constitution Party - Share few social values and some economic values.
Socialist Party - Share some social values and few economic values.

I obviously prefer the Libertarian Party over everyone else, but I also obviously would very much not want to see a Socialist or Constitutionalist win. Depending on whether I value social or economic values higher, I might prefer either the Greens or the Republicans, or I may compromise and choose the Democrats. However, any would be solid choices over the Socialists and Constitutionalists. Therefore, I might want to vote thus:

[FOR] - Libertarian
[NEUTRAL] - Democratic
[NEUTRAL] - Republican
[NEUTRAL] - Green
[AGAINST] - Constitution
[AGAINST] - Socialist

Then, my vote might be counted as +1 for the Libertarian Party and -1 each for the Constitution and Socialist Parties. Extrapolate this across an electorate of 25,000 and you might get something resembling the following:

Libertarian: 10,000 FOR/11,000 NEUTRAL/4,000 AGAINST = 6,000
Democratic: 12,000 FOR/6,000 NEUTRAL/7,000 AGAINST = 5,000
Republican: 13,000 FOR/3,000 NEUTRAL/9,000 AGAINST = 4,000
Green: 11,000 FOR/5,000 NEUTRAL/9,000 AGAINST = 2,000
Constitution: 5,000 FOR/13,000 NEUTRAL/7,000 AGAINST = -2,000
Socialist: 6,000 FOR/10,000 NEUTRAL/9,000 AGAINST = -3,000

In which the Libertarian candidate, who has the greatest majority of approval votes over disapproval votes, wins the election.

Since I think this method has some level of support, I wonder if it should be included in the article. Cheers! —Cuivi é nen, Sunday, 19 March 2006 @ 19:16 (UTC)

This is equivalent to Range voting. KVenzke 18:12, 6 June 2006 (UTC)

On Wikipedia Community, a democratic form to enlist candidates

(context: on Wikipedia we have Elections for the Board of Trustees)

Approval voting is a is a voting system used for elections, where a "list of candidates" must be pre-defined. If the list have "democratic origins", the election by approval voting will show a democratic result. It is a perfect system for wikipedia community...

But what is a "candidate" in a direct democracy? People that wants power? Marketing? Volunteers? Whant help others? It was good if I, or you, or your friends, be a candidate? (And how many candidates on the list to be a democratic list?)

We have on Wikipedia community a exceptional opportunity to do a "more truth" democracy... the goal of this community is to colaborate on articles, and through articles we know people. Only on this "known people" we can vote (!). Not names or promises on a "Elections list of self-candidates".

We haven't time to investigate another people, only ours "known people", and this is a good principle: colaborators voting on a colaborator from (and only from) the articles where he/she was colaborate. From this first ease voting we can produce a very surprised and democratic list of potenctial candidates... and invit the "best approval" of then to become candidates.

Operationally it is possible:

Principles:
1. Truth democracy need also to generate, all time, new and truth candidates.
2. Only about your "known (local) people" we can do a truth vote.
Possible wiki process to generate renewed and democratic list of candidates:
1. Each colaborator can votate on N (1, 2 or 3) colaborators from (and only from) M (2*N or 3*N not very more) articles where he/she was colaborate.
2. We will have a "per article", "per language" or "per etc." (level) elected ones. They will be invited to become "per level" candidates.
3. The confirmed candidates will go to compose the "list of candidates".

-- User:krauss (please sorry my english if necessary)

deleted orphan

Found this hanging around

" However strategy issues of candidate list order if voters are not fully aware and reflective of the full set of candidates before any votes are cast.) "

Looks like an editing remnant?

"Generalized Spectral Analysis for Large Sets of Approval Voting Data Article" isn't about approval voting in the sense of this page (I think) and should be removed

Unless I am misunderstanding, the article "Generalized Spectral Analysis for Large Sets of Approval Voting Data Article" by David Thomas Uminsky in the links seems to be about the case where there are only two "candidates", and only one of them will be selected. The author calls this "approval voting", but I don't think he means the same thing as is usually meant by the term; in his case, I believe approval voting is the same as plurality voting.

So, I think that link should be removed.

Bayle Shanks 07:58, 2 June 2007 (UTC)

You are right. Removed. --R.H. (talk) 22:12, 19 August 2009 (UTC)

"misleading changes to section on tactical voting"

I see that my sourced revisions were reverted by someone who didn't have the time to explain why they are so "misleading" and who in the process deleted two peer-reviewed sources. Of course I am reverting back and asking for an explanation. ←Ben^B4 02:20, 22 September 2007 (UTC)

Thanks. ←Ben^B4 06:17, 26 September 2007 (UTC)

You still do not understand my point. I have always been talking about a single seat election. The "expected outcome" is not generally the most likely winner. I am using the word "expected" in the sense of probability theory — an AVERAGE over the candidates with the utility of each candidate multiplied (weighted) by his probability of victory. If only two candidates have a substantial probability of victory, that average will usually be somewhere in between the utility of one of them winning and the utility of the other winning. Your wording over-simplifies the method.

Another issue — properly done, approval voting would involve only one round of voting. If two or more candidates receive the maximum number of votes, then one of them would be selected at random. There would be no primaries and no run-offs.

Approval is relative. So if there were only two candidates and you like both, you should only approve the one you like more. If you dislike both, then you should approve the one you dislike least. JRSpriggs 06:31, 26 September 2007 (UTC)

You are right that I do not understand. In terms of "expected outcome," could you please describe a situation where it is different than the most likely winner? ←Ben^B4 07:09, 26 September 2007 (UTC)

Suppose there are three candidates, as follows:

candidate, probability of victory, utility of victory

Alice, 1/2, +500.00 US$

Bob, 1/3, 0.00 US$

Charlie, 1/6, -2100.00 US$

In this scenario, Alice is the most likely winner; and Bob is second most likely. Using the simplified tactic, one would vote for Alice only. However, the average utility is

+500.00 * 1/2 + 0.00 * 1/3 + -2100.00 * 1/6 = 250.00 - 350.00 = -100.00

This expected outcome is less than Bob's utility. So you should actually vote for both Alice and Bob. That is, Charlie is so bad relative to the difference between Alice and Bob that the necessity of defeating Charlie out weights the preference for Alice over Bob. That is, you should risk getting Bob rather than Alice to be as certain as you can be of avoiding Charlie. JRSpriggs 03:27, 27 September 2007 (UTC)

That requires that the voter knows the expected utility instead of having just pairwise preferences. I understand now; thank you. ←Ben^B4 08:44, 27 September 2007 (UTC)

Thanks for fixing it. JRSpriggs 00:09, 28 September 2007 (UTC)

There is a problem with the section on Tactical Voting. It is questionable whether or not setting an Approval cutoff is a "tactic." The section implies that there is some insincerity involved. A close examination of what "sincerity" means, and its application to Approval Voting, shows that no Approval cutoff is clearly insincere, it is simply a decision, unless there truly is no preference. That is, there is no absolute standard for a "sincere" approval cutoff, any level of expected satisfaction could be chosen by the voter, but it makes sense that a voter will understand that *strategy*, if the voter wants his or her vote to have an effect, will involve setting the Approval cutoff between the two frontrunners. If there are three frontrunners, the analysis gets more complex, but the point is that there is no definition of "sincere" in setting the approval cutoff; basically, a clearly "insincere vote" in Approval, one which reverses preference, makes no sense, and I understand that Brams deliberately designed Approval to have this characteristic.

Ranked methods and Rating methods (Approval being the simplest Rating method, I call it Range 1) have truly different meanings for "Strategic." In Ranked methods, particularly if equal ranking is not allowed, there is only one kind of "strategic voting", which is preference reversal. In Approval, there is only the "insincerity" of equally ranking two candidates when you actually have a preference between them. But, since the method forces this if there are three candidates, unless you have no unique preference order for them, it cannot be said to be insincere. (With three candidates and two ratings, you have to have the same rating for two candidates! -- unless abstentions are excluded, in a Yes/No Approval election, which is quite contrary to precedent.

This ambiguity in the meaning of "strategic voting" is used by opponents of Approval Voting for polemic effect. And that spin is in this article.

There is indeed Approval "Voting strategy," just as there is the same for *any* voting method. A "voting strategy" is a method for converting internal preferences and preference strengths into votes or voting patterns, in order to maximize expected outcome for the voter. There is no avoding strategy, for a voter who wishes to take responsibility for the effects of his or her actions, and a vote is an action, not an expression of sentiment. The whole application of the term "sincerity" to it is problematic, and it gets really dicey when unexamined assumptions are incorporated into the discussion. We see this more intensely with argument about Range Voting. There is a special usage of "sincere" with Range Voting, but, problem is, there is no way that it is defined universally. Range Voting is theoretically optimal with all-"sincere" voters, but to be truly optimal, these voters must vote absolute utilities. Normalizing them to a particular scale spanned by the candidate set causes the performance to decline, because it equates what may be a very strong preference range for one voter with a very weak one with another. There is no defined "sincere" vote. Because of this, Range appears to fail Independence from Irrelevant Alternatives, but this is only true for normalized votes, not for "sincere absolute" ones. The presence of an irrelevant alternative can cause Range ratings to shift only because of normalization. It's a *very* complex issue.

I find it much, much simpler to consider votes as ... votes. Actions. You vote for a candidate if you want to help that candidate win. If you don't want to help the candidate to win, you don't vote for the candidate. It's that simple. Whether or not you want to help a candidate to win depends on the other candidates present, because you may be limited in your *real* choices; i.e., the ballot may be presenting you with irrelevant alternatives. You decide, not some ballot designer who tells you to "Vote for candidates you approve," where to put your weight. Because only one candidate can win, you can put your weight on more than one, because only one of your votes *at most* will actually count, the others are moot. That is why Approval has also been called Alternative Vote.

And then to understand Range, we simply consider that voters are allowed fractional votes. Reform sequence:

Plurality. Vote for one, discard ballots with votes for more than one. Approval: Count all the votes, i.e., don't discard such ballots, count them. Range: allow fractional votes. --- or --- use ranks and pairwise analysis of some kind.

Range is the *simplest* step in increasing voter freedom beyond Approval. Use of ranks, unless preference strength information is incorporated, is far more complex in its implications and strategies. Abd 03:48, 6 October 2007 (UTC)

If you think that BenB4 or someone else misrepresented approval voting as more vulnerable to distortion by insincerity than it actually is, then please correct the article. Personally, I find it difficult to define what such insincere (sophisticated) voting would be with sufficient clarity to make a definitive determination of what effect the use of approval voting would have on it. JRSpriggs 21:28, 6 October 2007 (UTC)

Approval voting is immune to lying- lying about one's preferences on the ballot will only hurt the person voting. However, it is VERY dependent on where you draw the line between the 'approved' and 'un-approved' candidates. Because of that, it requires as much strategy as a simple first past the post voting system, except that in approval voting, you can also vote for everyone you like more than the person you'd support in a first past the post system.Paladinwannabe2 15:52, 11 October 2007 (UTC)

Count All the Votes

I started working on Instant Runoff Voting recently, and ran into a hornet's nest of sock puppets and an anonymous IP multiple-reverter who turned out to be Rob Richie, Executive Director of FairVote. (See Talk:Instant-runoff_voting for a description of what happened.) Eventually, my actions, and their escalating response, led to their banning from Wikipedia. I see that one of the sock puppets was very active here, this was User:BenB4. His edits were contrary to Wikipedia policy, in that he was violating Wikipedia rules by participating, since he had been banned under other names. I mention this because his edits are fair game for reverting, which normally should be reserved for vandalism. One might note that he used reverts freely in dealing with what he didn't agree with....

But my occasion for being here now is just to begin a discussion of the name for Approval Voting, and the implementation of the method, involved in Count All the Votes. We get Approval Voting, in effect, if we simply start counting all the votes on a standard Plurality ballot, all that is needed is to change the instructions, and to set the counting process as if it were a multiwinner top-N election. This could involve as little as the deletion of a sentence or two from the election code of a state.

It's simple, and it's clean. It's arguably not the *best* method, though some do think that Approval is, in fact, the best, and it is a reasonable contention. But that's not the point. Count All the Votes is proposed as a simple, nearly cost-free reform that does not preclude other reforms. It is a simple increase of voter freedom, and it solves the first-order spoiler effect, and continues to perform well, unlike IRV, if more than two candidates are in reach of winning. Discussion of this may possibly point to some shifts in how the Approval Voting page is presented. Approval Voting really is that simple: just stop discarding ballots merely because a voter gave two alternative winners instead of one.

A great deal of smoke has been emitted over the decision in Minnesota, Brown v. Smallwood, which, according to IRV advocates, rejected Bucklin Voting because it, as the ranks collapsed if there was no majority winner yet, became what we now call Approval Voting, and that the court considered that this was a violation of one-person, one-vote. That is a warped interpretation of the decision. The court actually made it clear, repeating it in a reconsideration, that it was the existence of alternate votes, allegedly unfair to those who preferred only one, and it seems quite likely that this court would have rejected IRV as well; there is legal opinion that this is the case, and a challenge in Minnesota to IRV is practically certain to happen. I think that Approval supporters may want to file a friend of the court brief in that case, because it would be a mistake if the appeal were based solely on the red herring of alleged multiple votes. There was a very cogent dissenting opinion in Brown v. Smallwood that explicitly denied any violation of one-person, one-vote, and it is clear that the majority of legal opinion at the time was that there was no such violation.

Brown v. Smallwood was a remarkable decision, in which the court majority cited a precedent and then immediately contradicted it without showing any consciousness that they were doing so. My opinion is that the precedent will be reversed if challenged, one point on which I agree with FairVote Abd 02:57, 6 October 2007 (UTC)

Sock Activity here / Questions of "Strategy"

I made some changes to the article yesterday, and I realize that some aspects of them might be controversial. They were taken out by a user, Special:Contributions/VSJA. That was a single-purpose account SPA created at the end of September, and only used here for three edits, including two on what I added, though one was harmless and even useful. Given that this article was one being managed by the sock Special:Contributions/BenB4 it's not surprising that further incarnations of this sock would continue to attempt to function here. One of this set of socks attempted to remove his account from Wikipedia, but goofed and did it from anonymous IP, thus revealing his IP; then he had reason to ask for removal of the talk page there, which happened, but not before I saved a copy. It looks like VSJA didn't make that mistake. He creates an SPA, makes a couple of hits, then drops the account.

Fine. Saves me the trouble of reporting.

As to substance, I'm certainly willing to discuss this. Approval Voting was actually designed to be strategy-free, and this is quite displeasing to some, and they managed to get papers published about the alleged vulnerability of Approval Voting to strategy.

It's a linguistic trick. With ranked methods, tactical voting refers to preference reversal, which is clearly insincere. In Approval, a voter effectively sets an approval cutoff. The voter can set that anywhere, there is no standard for what is "sincere." What would clearly be insincere would be preference reversal, where the voter approves a candidate less preferred than one he does not approve, which is silly. Brams was right. Approval is strategy-free, in the sense of never encouraging insincere voting. Does this mean that there is no "strategy" to it? Of course not!

A voter who has some idea about the likely outcome of the election can predict, commonly, that the winner will come from a set of two, at most. Certainly this is true in the U.S. for partisan elections! When there are many such candidates, nevertheless the voter may have some idea of who some of the top possibilities are. But with two it is simple: pick the favorite and set the approval cutoff between that one and the other. And then use that to determine all the votes. This is well-known as a recommended Approval strategy. Certainly there is the possible but more complex application of game theory to the problem, but game theory will only suggest a more accurate position for the Approval cutoff; while this is a use of optimizing strategy, it is in no way offensive, Approval is not "vulnerable" to it, because that implies some kind of pathology, it is POV language, which is why I removed it.

But this argument, that Approval is vulnerable to tactical voting, is a key argument of FairVote in its campaign against Approval Voting, which is starting to get some traction, so it is attracting more opposition from the IRV advocates. Used to be they simply dismissed it as irrelevant. The socks who made a few edits here were much more active on the Instant Runoff Voting article, protecting the spin there. They got nuked. The Executive Director of Fairvote, Rob Richie, using anonymous IP to make repeated hacks at criticism of IRV legitimately inserted into the article, got his IP blocked. His associate and co-author, Terrill Bouricius, who registered an SPA to act in "protecting" the article, was also blocked, but, partly due to my intervention, his access was restored. We will see if I come to regret that....

I am certain: they, people associated with or supporting FairVote are now watching all my contributions looking for anything that reduces their spin. There are sincere supporters of Instant Runoff Voting, I am sure, who will restrain any excesses of mine; but I'm actually being pretty careful to leave my advocacy hat at the door when I'm editing articles, I put it back on sometimes on Talk. My POV causes me to notice opposite spin, to be sure. But that is how this should function: people with differing views come together to find agreement on basic facts and a good article.

Abd 04:16, 8 October 2007 (UTC)

What you are doing is deleting sources from the article -- peer-reviewed sources -- why are you doing that? The text you are replacing them with has no citation to any source. Is it just your personal opinion?

Do you know what happens when everyone in an approval election votes tactically? It sqeezes out the middle, moderate candidates -- going the opposite direction from the Borda count, and favoring radicals. You would know that if you read the sources instead of deleting them. VSJA 06:25, 8 October 2007 (UTC)

Well, I'm trying to engage with real editors, not socks. That an article full of unsubstantiated opinions gets published in a peer-reviewed journal does not make those opinions facts. The literature is not in agreement on this subject. Accordingly, we could take it out, or leave what we can agree upon. I addressed the issue above, but VSJA did not engage on the issue. To be sure, it will take a careful examination, and what I've claimed about the source articles will need to be looked at carefully. I'm trying to think of how and where to do that. There is a newly formed Election Methods Interest Group which is soliciting membership and participation from those interested in election methods,

But "we," in terms of the consensus that governs this article, in accordance with Wikipedia policies, does not include SPAs who register and immediately dive in, obviously fully engaged with an article, with the same tactics and interest as the banned BenB4. Sure, there is a small possibility that VSJA is ... what? A total newbie with a developed sense of how to maintain spin in an article, as we saw BenB4 doing with example after example? Small possibility. Very small. Much more likely, if he is *not* BenB4, he's a meat puppet. Election experts are invited; and I think they are coming. What *do* election methods experts agree upon? Isn't that an interesting question? I think we should find out.

Note that this article has now been flagged because of a contradiction between what [User:VSJA|VSJA]] put back and what is in the Tactical Voting article. To be sure -- and for full disclosure -- I edited that article too, which had totally unsubstantiated claims in it. However, I'm also talking about what is going on here in the Wikipedia Election Methods world elsewhere, places where election methods experts hang out, and *not* merely those who agree with me. I'm seeking *real* participation by those knowledgeable on this subject, and not just here, but elsewhere. In the absence of a way that election methods experts can express some consensus, organizations like FairVote can pretend that they have the support of "experts," when, in fact, it is only a few. If we get the experts together and they agree to something, and that isn't my personal opinion, I'll definitely stand aside. Until then, VSJA, you are free to complain. You know the system.

(Notice, by the way, that VSJA doesn't have a user or talk page, I can tell by the link color. Perhaps, if he's real and not a sock puppet, he'd tell us something about himself, on his user page. Does he have experience with the study of election methods, or is he an activist?) Abd 05:14, 9 October 2007 (UTC)

I will do so. Where is the evidence that "the literature is not in agreement"? And what exactly are your qualifications on the subject? VSJA 15:00, 9 October 2007 (UTC)

The problem is because there is disagreement on what constitutes 'tactical voting'. Wikipedia has tactical voting defined as 'voting insincerely' instead of 'voting strategically'. Approval voting, like IRV, is susceptible to 'voting strategically'- in fact, it requires voting strategically in more cases than IRV does. Unlike IRV, though, Approval voting NEVER encourages ranking one candidate above another insincerely. According to our currently accepted definition of tactical voting, Approval voting never involves tactical voting. —Preceding unsigned comment added by Paladinwannabe2 (talk • contribs) 20:32, 11 October 2007 (UTC)

Does Approval Voting satisfy the Majority Criterion?

User:Tbouricius added a comment that Approval Voting fails the Majority criterion. This is a common opinion, however if we look at the Criterion as stated in the article, we see:

The majority criterion is a voting system criterion, used to objectively compare voting systems. As applied to ranked ballots, the criterion states that if a majority of voters ranks a given candidate higher than every other candidate, then that candidate should win. For non-ranked ballots (e.g. Approval voting or range voting) it can be expressed as follows: "If more than half of the voters give candidate X a higher rating than any other candidate, the winner should be candidate X."^[1]

We can consider Approval Voting as a ranked system which allows equal ranking. If a voter marks for a candidate, and marks for no others, this voter has expressed an exclusive preference for that candidate, this is ranking that candidate "higher than every other candidate." And, as with Plurality, which is really an almost identical system, only with the *added* restriction that only one vote is allowed, if a majority so ranks a candidate, that candidate must prevail.

In Approval, the voter, however, may choose to rank a candidate *set* higher than every non-member of the set. If a majority of voters so ranks such a set, the winner of the election must come from that set.

Now, you will find in many lists the claim that Approval does not satisfy the Majority Criterion. If you follow the reference in this article, however, you will find that James Armytage-Green has also stated that Approval does not satisfy the criterion, but he also notes that there is, to him, an ambiguity in the definition, he is thinking of unexpressed preferences.

We can say, yes, that it can be true that a majority of voters prefer A to B, but if enough of those voters *also* vote for another candidate, that other candidate can win. But this is not the Majority Criterion. Plurality is generally considered to pass the Criterion. However, a majority of voters could prefer A to B, but, in an election against C, they could think that only B has a chance of winning, so they vote for B. Thus, if we based our analysis of compliance on unexpressed preferences, Plurality cannot satisfy the Criterion, because it is possible for this failure to occur.

How likely is it? Opponents of Approval Voting are fond of claiming that Approval does not satisfy "majority rule," though majority rule is really a completely different principle, one which would, for example, suggest that a winner of an election should never represent a candidate opposed by a majority, which IRV, as described in the Instant Runoff Voting article, can certainly do, whereas the method described in Robert's Rules as "preferential voting," does continue to consider all ballots in determining "majority," and thus does not determine a plurality winner unless the specific election rules so allow. (Robert's Rules, by default, does not allow an election by plurality.)

In a two-party system, by definition, most voters are supporters of two major parties. These voters have no reason to do anything other than bullet-vote, unless perhaps they want to indicate some kind of support for a third party, they might give some additional approvals for that. But those who favor a third party, these will be the ones who will vote multiply. It might be only a few percent, but it's enough to deal with the spoiler effect. It would be *highly* unusual to see more than one candidate getting a majority, and even more unusual to see the winner be one not actually preferred by the majority. But, of course, it could happen. Nevertheless, this would be a good election outcome, one which receives the consent of the largest majority of voters!

I removed the claim. It's controversial, if it is going to be in there, the statement should be qualified.

Abd 06:01, 9 October 2007 (UTC)

This is incredible...Approval/Range Voting advocates appear to have distorted the Wikipedia page re-defining the universally accepted definition of the "Majority Criterion" as it can be found in any elections systems text book, in order to make a neat little exception for Approval?Range Voting. All political scientists who study voting systems agree that Approval Voting DOES NOT pass the Majority Croterion, and Wikipedia should not miss-represent that fact. This is a terrible example of inserting POV bias.

Here is an example of how the Majority Critereion is normally defined (this quote is from by James Green-Armytage from Antioch College)

"Majority criterion (MC): If more than half of the voters rank candidate X over every other candidate, then the winner should be candidate X.

Some methods that fail MC: approval [*], ratings summation, Borda"

Wikipedia should reflect scholarly consensus not POV of zealots.

I will undo this edit in both the Approval Article and the Majority Criterion Article.

Tbouricius 14:44, 9 October 2007 (UTC)

As to zealots, indeed. Tbouricius is an active proponent of Instant Runoff Voting, on the staff of FairVote. We have welcomed him here, because his point of view is important if we are to have true consensus on articles. But it appears that he does not understand that he cannot impose his opinions on all of us without discussion. The extended definition also came from James Green-Armytage. Now, I've discussed this very issue at some length with Mr. Green-Armytage, and he is aware that there is a problem: the definition of the criterion, as stated, depends on a meaning of "rank" that is something other than what is on the ballots. When methods other than Approval are being considered, the ballot is ranked, so the meaning of "rank" is unambiguous: it would be preposterous to insist that, say, a Condorcet method fails on the basis that the voters might have one preference order and vote another. "Rank" refers to what is on the ballot itself, not to the mental state of the voter. But with Approval, there are only two ranks: Approved and Not Approved, or Yes and No, with the No being assumed if there is no Yes. So how can we read this: "if the voters .... rank a candidate?"

What the analysis have done who claim that Approval fails the criterion is to take a criterion designed for ranked methods and apply it to Approval, which is in a different class of methods, though it does have two ranks, just like Plurality. Now, suppose that somehow we define rank to include hidden preferences, i.e., the voter actually prefers A to B but votes A = B. Yes, then, the "preference" of the voter, being in the majority, can fail to be elected. However, is this not also true of Plurality? For if a voter has a preference for A over all other candidates, but, say, does not believe that A has a chance, and perhaps a majority of voters think this way, they may vote instead for B and B wins. There is no guarantee of the victory of a majority preference unless the majority votes in such a manner as to express the preference. If they abstain, they have no guarantee! And this is what they do when they equally rank another candidate top; they no longer have the guarantee. They have not *expressed* an exclusive preference.

Is there a scholarly consensus on this point? I say not. Prove that there is. There is a way you can, Mr. Bouricius. Join the Election Methods Interest Group [1] and participate, for the goal of that group is, in fact, in part, to measure the actual consensus of experts as well as of others interested in election methods. I have no idea if the community of experts would consider Mr. Bouricius as an expert on voting methods, but they might. And they might consider me so as well, though I have no academic credentials. It's not up to me, it's up to the community of experts.

Mr Bouricius indicated his intention to edit the Majority Criterion article. He did so, and made a major blunder. But I'll point that out in Talk for that page when I correct his edit. Approval satisfies the Majority Criterion as written. Abd 06:02, 10 October 2007 (UTC)

If the majority criterion is "If a majority of voters ranks a given candidate higher than every other candidate, then that candidate should win.", then approval voting will not always satisfy it. However, I would question whether the majority criterion is actually a good criterion. It ignores the intensity of voters' preferences. If only 51% of the voters have a slight preference for Alice over Bob but 49% have a strong preference for Bob over Alice, then it seems reasonable to me for Bob to be elected rather than Alice. The difficulty, as always, is how to measure the intensity of voters' preferences reliably, that is, in a way that does not allow them to cheat by exaggerating. JRSpriggs 00:45, 10 October 2007 (UTC)

JRSpriggs, well-meaning as he is, apparently falls into the common error of considering hidden, unexpressed preferences, which no election method can incorporate in its analysis. Approval is really a Plurality method, only with equal ranking allowed; equal ranking does not deprive the majority of any power, except as the majority consents, explicitly, by voting for more than one. If Approval fails, so does Plurality, I don't think it is possible to have it both ways without an extraordinarily complex Criterion.

Whether or not the MC is a wise criterion to depend upon is an entirely different matter. It is easy to show than any optimal election method must fail MC, which is, essentially, a proof that Approval is not an optimal election method! It is merely quite a good one, compared to Plurality, and, yes, IRV. It certainly is cheaper than IRV! But if ideal method is what we want, we can then move on from Approval. It is merely a very good, very cheap, first step. Just Count All the Votes!

It should be known that I have discussed this question with a whole series of election methods experts, and there is a consensus on one point: the Criterion is not worded in such a way as to make its meaning, when applied to Approval Voting, clear. And I've seen various attempts to fix this, and, from my point of view, they all have failed, they lead to one absurdity or another. Maybe I should write an article about this. If I could get it published, peer-reviewed, and I think I could, then I'd be a published author in this field! Could I then quote myself here and tell the rest of you to stuff it? I think not.... Abd 06:02, 10 October 2007 (UTC)

The Majority Criterion doesn't apply to Approval Voting, because Approval voting is NOT A RANKED BALLOT. A ranked ballot implies an ordering of all the candidates, and approval voting specifically does not order all the candidates, it merely has a yes/no option for each one. We cannot 'consider Approval Voting as a ranked system which allows equal ranking' any more than we can 'consider a sphere to be a circle that goes beyond the plane'. Mathematical properties of circles don't all apply to spheres- you can't talk about chord (geometry) on a sphere, for instance. Bickering about whether Approval voting passes Majority Criterion is like debating if a perpendicular bisector of a sphere's chord passes through the center.

Paladinwannabe2 14:56, 11 October 2007 (UTC)

For an even better example, we could say Plurality voting is a 'ranked system which allows equal ranking'- certainly all the candidates other than the one you vote for are ranked equally. Would you argue that Plurality voting passes the majority criterion? If you say that Plurality voting fails the majority criterion, you're also arguing that Approval voting fails, because to argue plurality voting fails you have to rely on 'unexpressed preferences'. Paladinwannabe2 16:47, 11 October 2007 (UTC)

There is a claim made by Approval (and Range voting) advocates that the system is exempt from Arrow's Impossibility Theorem, and manages to side-step such tests as the Majority Criterion, because it is not a ranked voting method. I do not believe this assertion is supportable. For now let me limit my point to the Majority Criterion under discussion.

The majority criterion sets out a hypothetical scenario of voter preferences and then tests different voting methods in that unique scenario. The scenario imagined is one in which voters DO have clear preferences, and in this particular election a majority of voters consider one particular candidate to be the absolute best choice. This scenario exists PRIOR to deciding which voting method is to be used. The Majority Criterion test is simply; Is a given voting method certain to elect this candidate? If these voters in this scenario vote using certain methods (FPTP plurality, Condorcet, IRV, etc.) that particular candidate is certain to win, and thus they meet the majority criterion. If these voters vote using Approval (or Range, or Borda, etc.), that candidate may or may not win. Thus by definition, Approval fails the majority criterion.

One can argue whether the majority criterion is a good or useful criterion, but there really is no debate among political scientists about whether Approval fails this particular test. The fact that a particular voting method does not allow voters to express details about their full set of preferences, does not mean that voting system cannot be tested for a specific scenario, and cannot exempt it from the criterion.

Tbouricius 14:38, 17 October 2007 (UTC)

I don't remember if I read this somewhere or not, but I considered the same example and same conclusion against approval voting. A majority may vote against their best interest in Approval (by overvoting with a compromise choice) if they don't have knowledge of their collective power. Of course plurality can fail this easy case as well by an overuse of strategic compromise when afraid their top choice can't win. (In contrast a top-two runoff will pass this test since a sincere vote in the first round can be made for a candidate who stands a good chance of winning.) Tom Ruen 18:17, 17 October 2007 (UTC)

Tom, There is an entire other realm of voting method evaluation dealing with the issue of tactical voting, but the majority criterion is not in that category. It is one of the criteria that compare systems assuming all voters vote sincerely. In that case FPTP plurality complies, because IF a majority favor one candidate, and sincerely vote for that candidate, that candidate must win. That is not necessarily the case with Approval, as you point out, even if all voters vote sincerely.

There has been a recent edit war on this article in the section about tactical voting, which I have not joined. For now I am limiting my involvement to correcting clearly miss-leading statements that contradict what all political science text books have to say about Approval Voting and the Majority Criterion.

Tbouricius 20:37, 18 October 2007 (UTC)

Arrow's Impossibility Theorem

I would like to point out that Approval voting violates Arrow's Theorem, specifically:
unrestricted domain or universality- "the social welfare function should account for all preferences among all voters to yield a unique and complete ranking of societal choices. Thus, the voting mechanism must account for all individual preferences, it must do so in a manner that results in a complete ranking of preferences for society, and it must deterministically provide the same ranking each time voters' preferences are presented the same way."
Perhaps whoever put that up meant it doesn't violate IIA? Paladinwannabe2 17:29, 19 October 2007 (UTC)

I think that the person who said that Arrow's theorem does not apply was referring to the fact that the theorem takes preference ranking as its input, but Approval voting does not begin with preference ranking. While some may say that everyone has a preference ranking guiding their choice of an actual vote, I would say that what actually guides people (at least rational people) is an assignment of utilities to the possible outcomes. This differs from preference ranking because it is quantitative, i.e. it has spacing in addition to order.

So while Arrow's theorem is a very interesting theoretical result, it has nothing to do with rational voting. JRSpriggs 21:02, 19 October 2007 (UTC)

Actually, the theorem says you can't have a voting system with preference ranking, IIA, non-dictatorship, and other criteria. Since Approval lacks a preference ranking, it doesn't matter if it passes everything else, it still fails Arrow's criteria, and therefore does not violate the theorem. Paladinwannabe2 21:20, 19 October 2007 (UTC)

^ Voting definitions and criteria

[1] Voting definitions and criteria

[1]