Talk:Single transferable vote/Archive 2

Archive 1

Archive 2

Archive 3

Archive 4

Archive 5

Images

We'd do really well with some good pictures of actual STV ballots to show what voting is like. An Australian ballot demonstrating group voting tickets, for instance, would round out a section nicely. Also, replacing that MS Paint picture that's in there currently would be nice :) (unsigned 21:18, 28 May 2005 User:Scott Ritchie)

I am not sure about ballot papers: these really belong in the preference voting page; even group tickets are a form of preference voting. --Henrygb 22:13, 31 May 2005 (UTC)

Why not have them here as well? Nothing prohibits us from using the same image in multiple articles (the one there is used that way already), and they certainly add information about what STV is like, and since afterall this is meant to be an encylopedic article on STV, that can only help make it better. Scott Ritchie 00:02, 1 Jun 2005 (UTC)

The article is fairly long as it is. And there is plenty more on preferential voting (e.g. the potential for confirming vote buying) which could also go here. But I don't really care. --Henrygb 19:55, 2 Jun 2005 (UTC)

I added one for the Aussie system. It's not a real ballot I don't think but it gets the idea across. Felix the Cassowary 03:56, 12 July 2005 (UTC)

Wasted votes

I am not convinced that we can say something like "the number of wasted votes must by necessity be less than a quota". Two examples, the first plausible, the second not:

• A one seat STV election where 60% of the first preferences go to one candidate: the quota is marginally over 50% of the votes, but the wasted votes are the 40% going to the other candidates plus the 20% of the total votes which the winner won but did not need to win on the first round, making virtually 60%.
• A ten seat STV election with eleven candidates: 1000 voters give a full preference listing for the first 10 candidates in various orders with the candidate Z in last place every time, while 1 voter puts the candidate Z first. The quota will be 1001/11+1=92, but the number of wasted votes will be 980 (those of the 1000 which would not have made any difference to the result if they had not been cast) plus 1 for the first preference for candidate Z.

Clearly you could define wasted votes so as to ignore unnecessary votes for winning candidates, but taking that as your definition would justify an STV system which did not transfer surpluses, leading to fewer wasted votes. --Henrygb 22:45, 28 July 2005 (UTC)

• • Hmm, I see what you're saying. I will note that in your second example, the candidates are winning not because they fulill quotas, but because all other candidates were eliminated. It is still true, however, that the number of remaining excess votes (a subset of wasted votes) will always be less than one quota. Scott Ritchie 23:29, 28 July 2005 (UTC)
    • I had imagined that in the second case, the surpluses would shuffle round the first 10 candidates until they each had 92 or more; if there had only been 20 votes for them together and 1 for Candidate Z then the surpluses would shuffle round until the first 10 each had 2 votes (or the quota if a smaller fraction). But it doesn't really matter. --Henrygb 19:55, 29 July 2005 (UTC)

In answer to just above, there would be no shuffling, or very little -- the bottom candidate, the one with one vote, would be dropped on the count immediately following any vote transfers arising from surpluses. This would automatically leave only the same number of candidates as seats so resolve the election. In answer to the first point about wasted votes, an election using preferential voting that elects only one is not STV -- it is Alternative voting AKA IRV.

The type of wasted voting which STV is said to mostly prevent is the waste of votes not finding a final home among those elected, not votes wasted by being taken by winners who did not need them compared to the next leading candidate. The former is commonly accomplished under STV. In Edmonton, Alberta's STV election for provincial representatives (MLAs) 2 percent of votes were rejected in part due to confusion about preferential voting, 3 percent were "exhausted" at the end of the 17 counts because they did not bear enough back-up preferences to be transferred pbeyond eliminated (or successful) candidates, and 8 percent belonged to the last eliminated candidate whose votes were not transferred as the sole remaining candidate took the last open seat. The total 17 percent compares well to 49 to 70 percent commonly wasted under first past the post, or even the 30 to 49 percent wasted under Alternative Voting.

US dates

Lo all - I've shoved the list of US cities that used STV off to a specific list page, though useful; I don't think it belonged in the article.

If anyone could scare up some dates to put on that page I'd be chuffed to bits.--Red Deathy 07:48, 28 September 2005 (UTC)

I don't have the dates right now, but I can tell you where to find them. See the chapter by Leon Weaver in Electoral Laws and their Political Consequences (eds. Grofman and Lijphart 1986).

Also, now that the use of STV around the world has its own page, I think it would better to put the 22 cities in that page and not have a separate page.

Trimming the article

I have a wee proposal - the article is getting a bit ragged - again - I suggest that we hive off |Issues and |Use Around the World to their own seperate articles - as we did with counting?

I think something needs to be done....--Red Deathy 15:32, 14 October 2005 (UTC)

I don't think you can remove Issues and stay NPOV. But Counting could be shrunk as here is another article, Vacancies could also largely move to that article. Other paragraphs which I personally feel have low useful info/length ratios (as of 22:54, 16 October 2005 (UTC)) are the first and second of Issues, the second of Proportionality, the second, third and fourth of Tactical voting, the third and fourth of Effects on factions and candidates, and the third of Voting system criteria. --Henrygb 22:54, 16 October 2005 (UTC)

I don't think you can remove Issues and stay NPOV - I don't think that's necessarilly true, if we leave what would be a short factual description of what STV is - perhaps trimming the remaining article carefully for contentious phrasings - and then flag up the link to the issues which could then themselves be improved considerably as a coherent article-in-themselves. The Voting and History sections, which would essentially be all that remains - are fairly balanced in themselves. It's just my feeling is that people will be more receptive to a series of short articles than to what appears to be one epic article. Anyway, I'll leave trying the look of anything till the back of this week, and we'll see how it goes...--Red Deathy 07:23, 17 October 2005 (UTC)

My personal view is that calling STV "proportional representation" verges on POV - it often produces non-proportional results and it is not designed to be a party-based system anyway; semi-proportional would be a better description. Move that out with Issues and I would be less concerned. I am not sure I would agree about a series of short articles being better. --Henrygb 08:51, 17 October 2005 (UTC)

STV is widely viewed as proportional representation, however. In New York, for instance, the system of STV was literally referred to as "proportional representation", rather than as simply STV or some other name for the voting system. You'll find similar usage in texts about Irish history as well. "Proportional representation" does not simply mean party-proportional, both in common usage and among voting theorists. STV does provide Droop proportionality (anyone meeting the droop quota will obtain a seat), which is a term relevant to PR and probably deserving of an article. Scott Ritchie 01:05, 21 October 2005 (UTC)

A note of caution: I'm not sure the previous hiving off of Counting Single Transferable Votes has worked too well; that article is still a mess, nobody is fixing it, and the section here is not really a summary of it. Having said that, Tactical Voting could well be moved there (possibly the page would be renamed to something like "technical details of STV"?) since the issues of tactics interplay with the counting method. In fact, all of Voting should be moved there and summarised here (keeping the example here, though we need better examples there).

Issues could get a subarticle each, but of course need to be at least summarised in the main article. (Vancancies could have a one-sentence summary, others possibly longer). STV around the world can easily be separated out, possibly taking History with it (USA and Canada sections are largely historical anyway). In summary, I agree that some sections have grown more than others, giving a lopsided feel to the article; subarticles could fix this, provided they are split into logical units, as self-contained as possible, and all signposted from this main article. Sounds like hard work. Joestynes 15:15, 17 October 2005 (UTC)

OK - Use of STV around the world seems to be an uncontroversial move, so I'll do that now, and we can see how it looks and think about if anything else needs trimming/moving. Everything can be easilly reverted so no harm should be done. --Red Deathy 07:11, 19 October 2005 (UTC)

Let's be careful not to overtrim the article, however. A very large amount of the "text" calculated as putting the article over the 32k warning level is simple markup language and comments (the example table, in particular, takes up more characters than several sections). Taking the current version of the article and copy/pasting the readable words into Office tells me that the current word count is about 35,000 characters in 5500 words, many of which are in the references/external links section. Unless it grows further, that should be condensed enough for now. Scott Ritchie 01:05, 21 October 2005 (UTC)

Yes, I think having moved use around teh world to a sensible sub artiucle, I think it looks more balanced, and about 5K words isn't too much. obviously, you can never stop trimming and condensing - so long as you can find simpler ways of saying the same thing. But I think further changes would need to be re-wordings rather than straight-up moves. --Red Deathy 08:08, 21 October 2005 (UTC)

On second thought, the cutting out of the use of STV around the world has been a bit abrupt - there was some interesting content in there. Perhaps it should be summarized in mainarticle format for the history section, rather than merely linked as a See Also? Perhaps we should even move the content there to History of STV, and include some of the juicier relevant bits in the main article's history section. At the very least, we should have a couple of sentences saying that STV is used in various places. Scott Ritchie 19:32, 22 October 2005 (UTC)

"Choice of Votes transferred" section

I removed the following text that creeped in after being on the front page:

---

Choice of votes transferred

A difficulty exists with the STV during the choice of which votes are to be transferred to other candidates after a particular candidate (candidate A) has been elected. For example suppose a candidate requires 1000 votes to be elected, and 2000 voters give their first preference to candidate A. If these 2000 voters are broken into two groups Ab and Ac depending on their second preference (Candidate B - Ab and Candidate C - Ac) and Ab has say 1200 members and Ac has 800 members. The difficulty now exists on the choice of which 1000 votes to transfer. If all votes transferred are in Ab then Candidate B has an advantage. If all votes in Ac are transferred then Candidate C has an advantage. In fact, this can often happen during the counting process if votes from particular regions of a constituency are counted before those of another constituency and the second preference in the first constituency is on average different from that in the second.

One might counter that it could be done proportionally. In this case, since 60% of Candidates A's voters were Ab and 40% Ac, then 60% of the transferred votes would go to Candidate B and 40% to Candidate C.

---

This is already covered in the article to some degree, and if expanded on should be in the counting single transferable votes article - all STV methods in use today either transfer them fractionally (such as Meek's method), or randomly (and therefore effectively fractionally). Scott Ritchie 23:44, 3 October 2005 (UTC)

---

I thought this section was quite good. If anything, I would have added to it, to say that the selection of which votes to transfer is an opportunity for election mischief and a disadvantage of STV. The examples used in the article are relatively simple (Ab and Ac), but in a more life-like complex example (Abdf, Abf, Aceg, Aeg, Age, ...) the transferral process is bound to disadvantage someone, even if done with perfectly honest intent. Bill Taylor 00:10, 4 October 2005

I see what you're saying, however we should probably mention something like this in a summary sentence then link to a relevant section in the Counting Single Transferable Votes article - it's somewhat of a minor point (and one that varies significantly per method since things like Meek's handle it differently), and the article is possibly too long and technical already. Scott Ritchie 21:17, 11 October 2005 (UTC)

The question arises in the example, where chocolate's surplus is transferred. The example has been contrived to have a natural solution, but the issue would still bear mentioning. G Colyer 22:07, 30 November 2005 (UTC)

Meek details

This statement is correct: (1) Meek's method also dynamically adjusts the quota during the counting process of the election, lowering it slightly to take account newly exhausted votes. In Meek's method, initially different ballots that express the same preferences after particular candidates are eliminated are weighted exactly the same - there is no penalty for ballots arriving at a candidate in an earlier round than others.

These two statements are incorrect. I deleted (2) and RedDeathy reinserted it as (3). (2) Meek's version contained the innovation that electors could rank preferences equally, but this option has not been used. (3) Meek also considered a varient [sic] on his system which would have allowed for equal preferences to be expressed.

When one talks about "equal preferences" they generally mean that one could give multiple candidates the same rank on a ballot. E.g. rank two candidates first, one candidate second, and three candidates third. Some STV methods can account for this, but Meek STV cannot. Further, Brian Meek did not discuss this in either of his two Voting Matters articles.

RedDeathy, could you please explain why you wrote (2) and (3)?

jeff 14:24, 6 December 2005 (UTC)

Lo Jeff, I inserted the following comment in the text: [!--(Meek also considered an alternative formulation in which voters would be allowed to indicate equal preference for some candidates instead of a strict ordering; we have not implemented this alternative.) See reference 1 Hill 1987 --] --Red Deathy 14:06, 7 December 2005 (UTC)

Malta Encore

Nevertheless, failures to produce partisan proportionality exactly analogous to the party affiliations of top choice candidates as occurs in list PR elections can be controversial. For example, in Malta in 1981 the party winning more than half the top preferences won less than half the seats, resulting in a constitutional crisis: see below.

My problem here is linking the first sentence to the second - in essence the Malta situation was STV behaving like a party list system, with simple droop quotas - if you'd used D'Hondt or any of the others with the same constituency boundaries you'd have got much the same result. OK, list constituencies use bigger electorates (because counting is much simpler) but the fact is that it isn't an explicitly STV problem, and I think that needs to be got across. --Red Deathy 12:43, 29 September 2005 (UTC)

You cannot have it both ways. Either it is impossible to judge how proportional STV is (then don't call it PR) or there are examples of controversy. If Malta had for example open list PR, then the natural response would be larger constituencies or even a national constituency. But STV makes larger constituencies harder as described at the end of the section Single_Transferable_Vote#District_size. Having 100-200 candidates from a single party would be hard even in Malta. --Henrygb 17:05, 29 September 2005 (UTC)

I'm not trying to have it both ways, I'm suggesting there is a problem of phraseology there, in that you are taking an example where STV behaves as a list system (effectively) where the weakness demonstrated would be one that any list system could suffer with equal constituency sizes. i.e. the controversy is not related to STV per se but constituency use. Yes, STV does favour smaller constituencies for reasons of complicatedness, and as you say, that is covered elsewhere in the article. OK, how about this - The practicalities of counting votes means relatively small constituencies are used in STV (see below). This means that where voting behaviour would be analogous to that in list PR elections can be controversial due to large remainders. or something along those lines but less cumbersome.--Red Deathy 07:20, 30 September 2005 (UTC)

When STV does not behave like a list system on some criterion, it is difficult to judge how proportional it is and you are left with personal belief. When it does, you say that it has the same flaws. But the fact is that anomalies in STV raise questions about STV and need to be pointed out. The US presidential election, 2000 had the same anomaly, and that raised questions about the electoral college even though other systems could produce the same problem. I simply think it is unreasonable to try to hid anomalies in apologetics.--Henrygb 19:39, 30 September 2005 (UTC)

My point is the anamoly - constituency sizes - is already adequately covered in the article, and that the pararaph as stands is not entirely accurate or clear, i.e. is badly phrased. I agree the Malta case needs examining, but I'm looking for a better way to explaining it. The Electoral Cllege issue is very different to STV's, there's not even the semblance of equality of votes in that system (deliberately so). third party re-prahsing suggestions welcome.--Red Deathy 09:47, 1 October 2005 (UTC)

My purpose is to say there have been cases of STV elections where the party with majority support has lost, and more frequently other cases where more popular parties have won fewer seats than less popular parties, and that this has led to controversy. I don't mind other paragraphs in the same or different sections saying this can happen with other systems too, or that sometimes STV does quite well, but I do mind what I see as POV apologetics in the same paragraph trying to hide or obscure the fact that STV does not always deliver what people expect from PR.--Henrygb 19:58, 1 October 2005 (UTC)

OK, how about: The practicalities of STV means relatively small constituencies are used compared to list-pr systems. This can mean that whilst the result in each constituency is proportional to the votes cast, the overall result can be disproptional. The Most striking example is in Malta in 1981 the party winning more than half the votes won less than half the seats, resulting in a constitutional crisis: see below. Now, I hope you're not accusing me of being POV here, I am genuinely trying to reach an accurate statement based on the evidence of the data you link to, to explain Malta without reference to constiuency sizes and remainders strikes me as misleading - we need to show how the anomolous result occurred. now, please, I'm trying to work towards an acceptable rewording, stating what changes I'd like to work toward, I'd appreciate some comments back rather than being accused of appologetics. I prefer to work this way than to make changes which get swicthed back and forth. .--Red Deathy 10:34, 2 October 2005 (UTC)

Your edit was reverted by Scott Ritchie [1] with the comment "Revert user:Red Deathy's undiscussed changes to previous version. Calling maltese results disproportional implies a POV interpretation of proportionality)". I think "undiscussed" may be pushing it a bit. --Henrygb 00:14, 4 October 2005 (UTC)

Whoops, sorry, didn't see this. A lot of edits came in that day (including some hefty vandalism). I'll comment tomorrow when I have more time. Scott Ritchie 04:08, 4 October 2005 (UTC)

Just to get in my reply to Scott first: 1) I'm sure we can all agree that 'proportionality' is tendential rather than binary (perhaps a point worth making somewhere). 2) Any deviation from strict proprtionality that affects the global outcome of the election is significant. 3) I think Henry is right to maintain that Malta is an example of a significant deviation from proprtionality, and this needs to be emphasised. 4) I also think that the behaviour in that election is effectively STV as an open-list system, so not analogous to a list-system. 5) That we need to explain teh Malta case rather than simply mention it (at the least we should reference teh fuller description below)--Red Deathy 11:36, 4 October 2005 (UTC)

Ok, I decided to be bold and just made a pretty huge edit to the proportionality section. I tried to keep everything in there intact, though I did rearrange things a lot and removed the Irish example. I think it's better now, while still conveying the same meaning, though it is possible I changed something pretty drastically. What do you think? Scott Ritchie 08:48, 15 December 2005 (UTC)

Changes

I've just made a few changes to the article. Basically I felt that the article was front-loaded with too much complex detail, too early on, and this made in inaccessable. So I've tried to make it so that it begins with a run down of the basics, then proceeds in later sections to go into details about ballot design and surplus transfer methods, etc. I've also put back in a brief guide to the basic mechanics of the count. I think this is too important to be entirely spun off to another article. I've also added some information here and there and made minor aesthetic changes.

Iota 18:28, 23 March 2006 (UTC)

The basic mechanics were there and were rather clear, until some changes seemed to completely reorder the article. Unfortunately, it seems to be in pretty bad shape now compared to when it hit the front page. Right now the lead section has fallen apart as well. Scott Ritchie 08:00, 28 April 2006 (UTC)

NP hard

Can you please summarise what NP-hard means. This is a wholly inappropriate use of jargon - one needs to read another whole article to deduce what it means. I am a politics student, and have studied politics as a personal interest for many years, but have never come across the application of NP-hard <or even any explanation of what it meant> in relation to STV.

If someone could explain that a bit better, I'd be v. grateful. (RM21 00:13, 14 July 2006 (UTC))

• I think the NP hard statements are a little over the top. Yes, I accept that it is possible to construct a situation where it is difficult to compute and that is what the proof does. But what is suggested here is that STV tactical voting is always NP-hard. That is the equivalent of saying that finding a prime factor of a large number is hard: yes it is for some, but not for 1000000000000000000000000000000000000000000000000000 or indeed any even number. Take for example the Knockiveagh 21 May 1997 election [2] (about 2/3 of the way down the page that is linked from a related page). It seems obvious to me that if the SDLP first preferences had been evenly split 805/804 rather than 946/661 then the party would have won two seats rather than one. Many parties try to manage their votes in this way especially in systems where full preferences are not expressed, and many voters are happy to follow guidance (as they are in SNTV where balancing the votes is even more important). --Henrygb 17:46, 10 July 2005 (UTC)
• Tactical voting is always NP-hard, however NP-hard problems are not always unsolvable. This is particularly true in the small districts of only 4 or 5 representatives common to many of the STV systems. Also, it is quite possible that erroneous assumptions about the viability of their strategy are being used by election campaigns. Also, tactical voting is easier in systems not using Meek's method, as stated in the article. Was Knockiveagh using that method in 97? Scott Ritchie 22:53, 10 July 2005 (UTC)

Trying to balance first preference votes between candidates from the same party is a standard vote management technique in many systems where there are substantial numbers of incomplete preferences. Getting it to work is sometimes difficult because some voters will not take instruction, but the mathematics of it are relatively simple:
If at the final stage there are three remaining candidates all under quota, two of which are from your party and one from another, and your two candidates have more than twice as votes between them as the other candidate, then if your candidates' support is evenly shared you will win both seats. If not then you may not. There is more likely to be balanced support in the final round if there is balanced support in the first round.
To stop that you would need a system which reduces the quota as votes drop off - does Meek's method do that? As far as I am aware, Meek's method is used in New Zealand and not in the other countries listed. If so, the article should make that clear. It is certainly not used in Northern Ireland, the Republic of Ireland or Malta. I don't know about Australia. --Henrygb 00:29, 11 July 2005 (UTC)

Meek's method does reduce quota as votes are exhausted.--TreyHarris 00:44, 11 July 2005 (UTC)

Indeed. I found this [3] --Henrygb 01:21, 11 July 2005 (UTC)

BTW, NP-complete problems always have many easy solutions, far far more than factoring, which is why NP-complete problems are not used for cryptography. Yes, the NP-hard part should be toned down. Its easy to imagine asituation where a significant amount of information proved that a specific voting strategy was optimal. JeffBurdges 22:27, 26 November 2005 (UTC)

Could you describe one - for a layman?--Red Deathy 08:43, 28 November 2005 (UTC)

On Karp's_21_NP-complete_problems, it says "many of the problems can be solved if restricted to special cases, or can be solved within any fixed percentage of the optimal result". A good example is the traveling salesman problem where many solutions are easy to find, just not all of them. Anyway, it might be true that "tactical voting in STV elections is vastly more difficult than with other commonly-used election methods", but just being NP-hard really does not support this claim at all. JeffBurdges 15:23, 29 November 2005 (UTC)

The best naive explination is: Some problems are hard and some are easy, but all problems have sub-problems inside them, as well as supper-problems containing them, and you can obviously glue problems together if you like. Said this way, it seem natural that "most" hard problems have big easy sub-problems. JeffBurdges 23:02, 29 November 2005 (UTC)

Ta, that's helpful, but (and sorry to be a pain) you say Its easy to imagine asituation where a significant amount of information proved that a specific voting strategy was optimal could you describe one? It'll help relate this discussion to the article and provide some cues for proper editing. --Red Deathy 09:18, 30 November 2005 (UTC)

A question: is Meek's Method of counting NP-Hard itself? It looks as if it could be: it uses approximation methods, gets highly complicated with extra candidates and seats, and even with computer counting in New Zealand, it led to weeks of delay in announcing the results. In that case tactical voting will be at least as hard. --Henrygb 12:18, 31 January 2006 (UTC)

No, definitely not. There's a difference between sounding complicated and taking an amount of time that grows exponentially, and there's a difference between that and the theoretical notion known as "NP-hard".

What "NP-hard" means is that if you find a way to solve it in a polynomial amount of time in any possible situation -- note that this can be any polynomial, even an unreasonably large one like n^1000000 -- then you solve a huge theoretical problem in CS at the same time. It has almost nothing to do with how long the votes actually take to count in real life, especially since many situations can be easy and yet the general problem can still be NP-hard. Or, the problem can be completely infeasible to solve but not be NP-hard because it doesn't have the theoretical properties.

Of course, we know that applying STV isn't completely infeasible. We have clear rules that tell us just how to do it, and people even manage to do it without the aid of computers in many situations. If the election results take a week to return, that's not theoretical computer science getting in your way, it's bureaucracy.

I'd say that computer science concepts that sound more relevant than they are and are so prone to be misinterpreted do not belong in this article. rspeer / ɹəədsɹ 04:31, 14 July 2006 (UTC)

Dead images

Where are they on the commons? That is the reason they were deleted. gren グレン 06:22, 16 August 2006 (UTC)

Current use

There remains no summary of current usage of STV in the article (pretty pathetic for a featured article). Instead, since the splitting off of content to History and use of the Single Transferable Vote, there is only a summary of the early history (only up to the end of the 19th century).

Someone needs to summarise the remainder of the content in the article I've linked to, and append to the section "History and current use" in this article.

In the meantime, I do not think this article should keep its featured status - for an article on STV it offers scant example of how it works in practice in places such as the Republic of Ireland (where it is used for presidential, local council, national parliament and European parliament elections - not to mention popular Irish usage for secret ballot votes in random groups, classrooms, organisations, etc.)

zoney ♣ talk 13:57, 17 August 2006 (UTC)

The article is already on the big side (perhaps too big) - the subject is now explored in three articles (Counting & History and use) - I think it reads better that way. This article is general, exploring the theoretical side with the practical put soemwhere else. Utlimately, if there is enough material, you could have STV in Country X articles, if you think there is sufficient scope...As it is, this article is pretty exchaustive.--Red Deathy 14:26, 17 August 2006 (UTC)

If it is getting too big, it needs other sections split off, and summaries of those sections provided. The article should give a *summary* of STV in country X - otherwise it is not comprehensive (especially if it doesn't properly act as parent article to its daughter articles). zoney ♣ talk 09:32, 18 August 2006 (UTC)

It lists the types of STV available, and cites where they are used (including citations about which systems are used in ireland) the focus is on the types of STV rather than on the places it is used. Real world examples should be (and are) scattered throughout the text to illustrate problems and benefits. Given that teh detailed work will be in the Politics of Somewheria page...--Red Deathy 16:14, 4 September 2006 (UTC)

Tactical voting

The "tactical voting" section is written with POV, and also has various impotant mistakes. I'd like some help cleaning it up.

I'm going to go through the current version and make comments.

The single transferable vote eliminates much of the reason for tactical voting.

This is not just POV, it's puffery. STV is less succeptible to tactical voting than what system, under what conditions?

Voters are "safe" ranking candidates they fear may not be elected, because their votes will be transferred after they are eliminated.

This is inaccurate, since it can easily cause a candidate to be eliminated for lack of first-round support.

Similarly, voters are also "safe" voting for a candidate they believe will receive overwhelming support, because their votes will then get reallocated to their next preference, though with less than the value of a full vote.

This can be accurate under the assumption that the voters for that candidate have similar preferences to the voter, but inaccurate otherwise. If two candidates can be elected out of (Perfect_Candidate, Far_Right, Far_Left) and everyone wants Perfect_Candidate to win, a tactical voter will vote for Far_Right or Far_Left first, since otherwise the voter is essentially losing a large fraction (in this case around 1/2) of his/her vote.

Tactical voting is chiefly accomplished in STV by making assumptions about the other voters.

This is true not just of STV but any votng system which depends on more than one vote. It should be stated more generally or dropped.

A preferred popular candidate can be assumed to win and thus ranked lower on a tactical voter's ballot, allowing the voter to give more weight in transfers to his second-choice candidates (and, implicitly, giving fellow supporters of the popular candidate less weight.) This is particularly effective in the older STV systems still used in many countries that prevent elected candidates from receiving additional votes, as in that case none of the tactical vote is diluted on the already winning candidate. Under such old systems, a voter may even rank a non-preferred candidate that is assumed to lose first in order to increase the chances of his vote arriving late. This method of tactical voting is much less effective in the New Zealand STV system using Meek's method, however, as votes receive the same fractional weighting regardless of when they arrive at the successful candidate.

This is one sentence about manipulatability and three of apologetics for STV, raising my concerns about POV.

Though still theoretically possible, figuring out how to successfully vote tactically in modern STV systems by exploiting the non-monotonicity in this way can be computationally difficult.

I don't believe that this is correct. Yes, it's NP-hard, but in most cases at least one tactical decision is obvious, even to the common voter. In fact, with less than (say) 10 viable candidates, the problem is almost trival. Further, tactical voting instructions can be determined centrally (as by a political party), which in addition to reducing the problem of computational difficulty (since, if needed, mainframe time can be used). This raises a question about division of power, especially if parties use tactical voting advice to advance secondary concerns not shared by all of their members.

This makes tactical voting in STV elections vastly more difficult than with other commonly-used election methods.

This should be dropped or referenced. There's no reason to assume a priori that other systems' manipulatablities are in NP \ NPC.

As a consequence, the difficulty of tactical voting in STV elections increases sharply as the number of voters, candidates, and winners increase. This gives an incentive for larger electoral districts other than their increased proportionality, since particularly small electoral districts may have few enough candidates to make tactical voting feasible.

I don't believe this is correct. The difficulty of tactical voting should be independant of the number of voters, although it depends strongly on the number of viable candidates and winners.

Of special note is that voters have a real incentive to list their preferences honestly in STV, as it is the best strategy for securing representation if tactical voting is either impractical or impossible.

This is true with almost all serious methods. It should be removed or generalized.

This is frequently the case in large electoral districts, as successful tactical voting (when possible) requires both nearly perfect information about how others are voting and the computation of a virtually unsolvable math problem.

This is (1) blatantly incorrect, see above, and (2) redundant. Little information is needed to achieve some effect with tactical voting, and the computational difficulties are quite easy with small numbers of candidates. It's true that (near-)perfect information is needed to have perfect tactical voting, but that's a much higher bar that has little practical relevance.

I don't want to come off as a critic of STV: I prefer it to most methods currently in use, in fact. However the article (or at least this section) comes off as advocacy more than encyclopedia. CRGreathouse (talk | contribs) 01:03, 21 August 2006 (UTC)

I’m thinking of rewriting this section (the one with POV notice) allong the following lines:

The single transferable vote eliminates much of the reason for tactical voting. Voters are "safe" ranking candidates they fear may not be elected, because their votes will be transferred after they are eliminated. Similarly, voters are also "safe" voting for a candidate they believe will receive overwhelming support, because their votes will then get reallocated to their next preference, though with less than the value of a full vote. However, the Gibbard-Satterthwaite theorem states that tactical voting is possible in any deterministic voting system where any candidate can win, and that STV is no exception.

Remove this paragraph – this is relevant to single winner methods/ bloc vote not proportional methods. The potential and effectiveness of tactical voting under all proportional systems is related to the size of districts in which proportionality is determined – tactical voting can generally only effect the outcome in 1 seat per district. The larger the districts the less effective it is.

Tactical voting is chiefly accomplished in STV by making assumptions about the other voters. A preferred popular candidate can be assumed to win and thus ranked lower on a tactical voter's ballot, allowing the voter to give more weight in transfers to his second-choice candidates (and, implicitly, giving fellow supporters of the popular candidate less weight.) This is particularly effective in the older STV systems still used in many countries that prevent elected candidates from receiving additional votes, as in that case none of the tactical vote is diluted on the already winning candidate. Under such old systems, a voter may even rank a non-preferred candidate that is assumed to lose first in order to increase the chances of his vote arriving late. This method of tactical voting is much less effective in the New Zealand STV system using Meek's method, however, as votes receive the same fractional weighting regardless of when they arrive at the successful candidate.

This frequently mentioned method of tactical voting applies only to systems that use the simple Gregorian transfer method (only votes from the last parcel transferred to an elected candidate are considered for transfer). This tactic could, in theory, work for systems like Cambridge, Massachusetts STV, Irish Republic Handcounting and the Newland-Britton family of systems (Newland-Britton Northern Ireland, ERS97). This tactic would be completely ineffective for inclusive Gregorian transfer (Australian Senate), weighted inclusive Gregorian transfer (BC-STV), Meek and Warren.

Provide example of how this form of tactical voting could work, add statement that if all voters vote for candidates they don’t believe will win in the hope that their votes will transfer and have greater weight in determining the outcome, the candidates who nobody believed could win WILL win. This strategy doesn’t work if everybody tries to use it. Provide real world example of attempt to use this strategy

Though still theoretically possible, figuring out how to successfully vote tactically in modern STV systems by exploiting the non-monotonicity in this way can be computationally difficult. It is NP-hard to determine whether there exists an insincere ballot preference that will elect a preferred candidate, even in an election for a single seat. This makes tactical voting in STV elections vastly more difficult than with other commonly-used election methods. Importantly, this resistance to manipulation is inherent to STV and does not depend on hopeful extraneous assumptions, such as the presumed difficulty of learning the preferences of other voters. Furthermore, it is also NP-hard to determine when an STV election has violated the monotonicity criterion, greatly reducing the likelihood that the electorate will know if even accidental tactical voting has occurred. As a consequence, the difficulty of tactical voting in STV elections increases sharply as the number of voters, candidates, and winners increase. This gives an incentive for larger electoral districts other than their increased proportionality, since particularly small electoral districts may have few enough candidates to make tactical voting feasible.

Providing an explanation of this form of tactical voting in terms of monotonicity (or the lack of it) or NP-hardness makes the explanation difficult to understand and overly technical. The commonest form of tactical voting (organised by parties, for example Sinn Fein in Belfast West for the last NI Assembly elections which involves non-monotonicity) is what could better be described as vote splitting or vote equalisation.

Example

Three parties A, B and C present the following candidates- A1, A2, B, C1 and C2. 1000 votes, 3 seats

All B voters prefer party C to party A. All C voters prefer party B to party A. All A voters prefer party B to party C.

Party A has 410 votes, Party B has 190 votes and party C has 400 votes.

A1 210

A2 200

B 190

C1 250

C2 150

C2 is eliminated first and the winning set is A1, B and C1.

If party C’s votes are split equally between its two candidates for example:

A1 210

A2 200

B 190

C1 200

C2 200

B is eliminated first and the winning set is A1, C1 and C2.

Describe with an example the form of tactically voting mentioned above, provide real world example, provide commentary on the problems with organising this sort of thing – it requires good information about voting intentions, cooperation of the voters, lots of organising, can go wrong if you don’t get it exactly right, the vote distribution between parties makes this sort of thing possible in only a minority of constutuencies.

Of special note is that voters have a real incentive to list their preferences honestly in STV, as it is the best strategy for securing representation if tactical voting is either impractical or impossible. This is frequently the case in large electoral districts, as successful tactical voting (when possible) requires both nearly perfect information about how others are voting and the computation of a virtually unsolvable math problem. Since tactical voting in STV works by effectively substituting one's own alternate preferences for transfers with other supporters of the same candidate, the effectiveness of tactical voting is greatly reduced when other supporters of preferred candidates have similar second-choice preferences. Although there is no way to completely prevent tactical voting by hiding support for preferred candidates, the tactical voter carries the significant danger of his assumptions about the popularity of his preferred candidate being wrong, risking his most preferred candidate losing because of his miscast tactical vote. This contrasts heavily with non-proportional, plurality-based systems, where there is both tremendous incentive and ability to vote tactically in order to avoid the spoiler effect.

Get rid of this paragraph – non-proportional systems generally provide greater incentives for tactical voting than proportional systems because the use of tactical voting generally can only effect one seat per district using proprtional methods.

Dgamble997 4 Sept 2006

Is you example actually about tactical voting - it's relevent but not tactical. Surely if you have people who support a party strongly enough to pursue voting to maximise its representation they are actually expressing their true preference. It would be tactical voting if they expressed a first preference for B in order to ensure, that, say, A is knocked out entirely or to ensure both C's win.--Red Deathy 16:02, 4 September 2006 (UTC)

I think the example is tactical voting, though not one of the more extreme versions. Regardless, I'm fine with Dgamble's suggestions in general. I think there should be some mention of Gibbard-Satterthwaite, and I don't like the line "Tactical voting is chiefly accomplished in STV by making assumptions about the other voters." because it's insufficiently general, but these are more quibbles than anything else. CRGreathouse (t | c) 16:12, 4 September 2006 (UTC)

I've always considered tactical voting to be voting differently to how you would sincerely vote in order to alter the result of an election. Voters spliting their votes equally between two candidates of a party when they would normally favour one candidate of that party in order to increase the party representation is IMHO a form of tactical voting.

Vote management strategy is one of the few (possibly only) potentially effective forms of tactical voting in STV. See article

http://www.tcd.ie/Political_Science/Staff/Michael.Marsh/LagunaBeach.pdf

Vote management is described on page 10

Dgamble997 5th Sept 2006

Precisely, my point was that for someone to be sufficiently committed to a party to vote in a managed way probably is expressing their true preference (they might not care about which candidate so long as it's their party, or if they did and still followed instructions, that suggests their preference is the party first) - I think calling it vote management might be more useful than tactical voting--Red Deathy 07:40, 6 September 2006 (UTC)

But there are two other clear versions of tactical voting seemingly unique to STV

• one is "first preferences for candidates who will be eliminated in the middle of the count" which works for non-Meek methods even when you really support a popular candidate first but want your transfers to count as much as possible
• another is the Australian group voting tickets (also used in Fiji) where parties tactically swap preferences for tactical rather than ideological reasons.

--Henrygb 09:09, 6 September 2006 (UTC)

Several good points. It is worth noting that group voting ticket and voting for a canadidate certain to be eliminated only work for certain STV systems. Group ticket voting only works if you have above and below the line voting and "first preferences for candidates who will be eliminated in the middle of the count" only works for single Gregorian transfer - not for Meek, Warren, BC-STV (inclusive weighted Gregorian transfer) or Australian Senate rules (inclusive Gregorian transfer). I'll bear all these points in mind.

Dgamble997 8th Sept 2006

New section completed. I've only been on wikipedia 2 weeks so the layout and references could possibly be done better.

Dgamble997 10th Sept 2006

References

The Voting Series

This page should be part of the voting series with the voting tag {{{voting}}} however, due to pictures and diagrams in the article, i struggled to place it correctly. Could someone please add the tag as best they can? Woodgreener 21:00, 6 January 2007 (UTC)