CZ Talk:Ombudsman/Archive 1/Election process

From Citizendium
Revision as of 16:04, 7 March 2012 by imported>Dan Nessett (→‎STV and Instant-runoff)
Jump to navigation Jump to search

Report of Election Observer 2010

Please note the report I made as election observer after the 2010 elections Gareth Leng 09:51, 1 February 2012 (UTC)

More specifically, here is that report. D. Matt Innis 15:31, 6 March 2012 (UTC)

Referendum on elections

So this is an instant-runoff system? The idea is to transfer votes from eliminated candidates? John Stephenson 13:17, 27 February 2012 (UTC)

Referendum A (election of unopposed candidates)

Please also note that Referendum A had slightly more support than Referendum B, and got a majority, but failed because it did not gain the two-thirds necessary to modify the Charter. However, there is clear support for the idea of allowing single candidates to be elected unopposed and something like this ought to be implemented to avoid endless by-elections. John Stephenson 13:17, 27 February 2012 (UTC)

Reservists

Another way to avoid frequent by-elections is to fill vacant seats by offering them to unsuccessful candidates. I do not think that having reservists is a good idea, however, because it is likely that in some cases these will be people whose candidacies attracted little support. Actually, allowing the Combined Council to make interim appointments until the next election might be best. John Stephenson 13:17, 27 February 2012 (UTC)

STV and Instant-runoff

I prefer the single-transferable vote and instant-runoff systems. Regarding STV, a proportional-representation system is the only way to ensure that the winners reflect the preferences of the voters and that votes are not wasted. Even if someone is 2nd or 3rd choice, they are still a candidate who the voter is satisfied with. Low 1st choices can, if necessary, be dealt with by imposing a minimum threshold such as the 25% in the referendum. STV also reduces the likelihood of a tie, since they can be broken with 2nd and subsequent preferences. With non-transferable votes, the only options are to re-run the elections or draw lots, which are truly last-resort options for STV.

The system is somewhat more complex than simple-majority voting but there is software to do the sums. The problems that occurred when STV was used in the first post-Charter elections were actually caused by the fact that people were allowed to contest two groups of seats and then give a preference, such that the order that names were removed affected the final results. But this was not a problem of STV itself.

I think I am the first elected member to get in without ever receiving an absolute majority (Hayford got the same number of votes as me, but he also got in previously under STV). I do not feel that I have a mandate, which is why I don't make casting votes on the EC. STV or another transferable system would have avoided this. John Stephenson 13:17, 27 February 2012 (UTC)

Writing here personally, not as a member of the MC, I oppose any voting system that requires software to analyze results. My objection is that such systems fail to provide transparency. If a voting system is so complex that a typical non-technical member can't determine the outcome from published voting information, then those members must trust an algorithm they neither understand nor are able to execute. This would have a chilling effect on election participation rates and would make it very difficult if not impossible for the election committee to adequately explain the election results.
In addition, the mandatory use of vote counting software implies requisite skills that would likely dissuade a large proportion of the community from serving on election committees. This would increase the difficultly of finding citizens willing and able to serve as committee members. Dan Nessett 19:40, 27 February 2012 (UTC)
Now that ballots are no longer to be destroyed soon after the count, the results are there should anyone demand a recount. I see no reason why an explanation could not be given of how votes were redistributed. Also, I believe that voter turnout has actually gone down since simple-majority voting has been used. As for people being dissuaded from acting as election committee participants, if someone is unwilling to participate because they don't want to get involved with software that is not actually that difficult to use, I really don't want them counting votes at all. What would such a person do in the event of a discrepancy even in a simple-majority contest? John Stephenson 07:08, 28 February 2012 (UTC)
Looking at the last few election results it would appear that the election committee could have counted them all on their fingers and toes, and the amount of digits they might need for such a task is getting fewer each time, so we are not discussing whether our new voting system would be so complex as to confound any of our current members, but instead wondering if with more members any new voting system might be better software-driven rather than limited by how many limbs the election committee might have.
If any concerns about software are about finding people able to use (probably simple) software to verify election results then don't worry, if we have enough voters to make software a necessity we will have enough citizens to find one or two not frightened of using it.
As for whether to use a preference system, of course we should. We've already seen the problems our current system has. It produced an alternate that the MC refused to allow to serve office. Their reasoning appeared to be that the alternate didn't get enough support? That being the case it would appear they used the STV system for judging suitability when useful. Although they did keep secret and destroy the results back in those days so it is hard to be sure.
The only reasoning I've heard against a new system is that it is either complex or un-American (that really is something that has been said on the forums), but if we attract enough members to make software a requirement there will almost certainly be a large portion of them that won't be worried by either of these arguments. David Finn 08:23, 28 February 2012 (UTC)
=> John Stephenson - Your arguments have several flaws. First, my point was someone should be able to determine the results "from published voting information." The raw ballots are never released for public display, since that would destroy the whole notion of a secret ballot. No matter how long ballots are retained, transparency requires the ability of non-technical members to determine the results from published information without the intervention of software, the operation of which they do not understand. Any voting system that cannot provide such transparency is flawed.
Second, what is simple to use is an objective judgment. What might be simple for someone with training in the sciences or engineering may not be simple for someone without such a background. Limiting election committee members to only those with technical skills introduces a tendency toward elitism and technocracy that could discourage those in the arts to participate in the project.
Third, even those with a technical background may have reservations about using software to determine the results of an election. I have a Ph.D. in computer science and have the skills necessary to understand how software works. However, taking the time to convince myself that a program correctly implements some non trivial vote tallying algorithm (and taking the time to learn and understand that algorithm) is not something I wish to do. I imagine many others feel the same way.
Fourth, there were 72 ballots cast in the charter referendum, 46 ballots cast in the October, 2010 election (which did not use simple majority voting), 40 ballots cast in the June, 2011 election, 14 ballots cast in the October, 2011 special election and 24 ballots cast in the December, 2011 election. Taking the special election as an outlier, participation has monotonically decreased over the past year and a half. If the use of majority voting is the cause of this, why was there a decrease in the October 2010 election? What evidence exists that use of simple majority voting has anything to do with this decrease? As far as I can see, no evidence supports this conclusion. Dan Nessett 19:40, 28 February 2012 (UTC)
The Charter vote was a yes/no referendum question so cannot be fairly compared with elections in which multiple candidates compete for several seats. Yes, people could also vote 'Revise', but as pointed out at the time, those votes were also to reject the Charter. Of course, there is no proof that the decline in turnout is anything to do with the voting system; that is a suspicion on my part. We know that a firm majority have voted to change the system. If so many are unhappy with the current system, that implies that some might have chosen not to participate in it. John Stephenson 12:54, 3 March 2012 (UTC)
I'm inclined to agree with your last comment.
I'm not sure what you mean precisely by verifiability from published information. The usual practice for STV elections in real life is to publish the totals for each stage. That is, you'd first list the numbers of first preferences for each candidate and then say whether anyone has been elected straight off. If so, you say how their surplus is redistributed. Otherwise, you say who's eliminated and how their votes are redistibuted. Would that be enough for you, or would you want a listing of all preferences on each ballot cast? Even the latter doesn't involve publishing the identity of the voter, or even correlating ballots in simultaneous elections.
STV procedure really isn't that complicated. I could easily work it out myself by hand, without software. So presumably could Peter. So very likely could David. And that's without any new training. You or anyone else here could do it if they want to learn. Peter Jackson 11:24, 1 March 2012 (UTC)

[unindent]

My concern is establishing a voting procedure that is so mysterious to non-technical members that they decide CZ isn't for them, If the algorithm of any voting method is easily implemented using pencil and paper ("easily" being defined as allowing someone without a technical background to execute it without a great deal of effort), then I would have no objection to it. While my view is we are attempting to fix something that isn't broken, if the majority of citizens want to use a voting method other than simple majority (which from the results of the referendum in the last election seems true) I will not oppose that. Dan Nessett 17:02, 1 March 2012 (UTC)

[added later] If STV voting satisfies the pencil and paper test, perhaps someone could explain how to to implement its algorithm with such implements using only published voting results (i.e., without having to divulge individual ballots). Dan Nessett 17:35, 1 March 2012 (UTC)
Let's distinguish between AV for a single post and STV proper for a batch of seats.
The former is really easy and shouldn't cause anyone any problems at all.
In the latter case, you'd probably prefer to use a pocket calculator to work out the proportions.
You still haven't said what you mean by "published voting results". Peter Jackson 18:42, 1 March 2012 (UTC)
For the purposes of this discussion, let's define "Published voting results" as those necessary to execute the algorithm. If a particular voting method requires the availability of individual ballots for this purpose, then I would say there is a significant conflict between it and the maintenance of secret ballot privacy. So, can you or someone else describe in detail how to use paper and pencil to implement both the AV and STV vote tallying methods? From this, we can determine whether the effort to implement those algorithms without software is within the capabilities of a non-technical member. Dan Nessett 19:05, 1 March 2012 (UTC)
Well, I'm still not clear what you mean, but I agree it's probably best to describe the method sufficiently here so everyone knows what we're talking about.

So let's start with AV, which is easier. You start by simply counting up the first preference votes. If someone has more than all the rest put together that's it. Otherwise you eliminate the person with least and pass on their votes to their next preferences. If they give no further preference then they no longer figure. If now someone has more than the remainder put together that's it. Repeat the process as many times as necessary. If a vote's next preference is for someone already eliminated you look for the next uneliminated preference. All really quite straightforward. The only problem you might get is if 2 candidates are equal bottom. Often it won't make a difference. If it does, what would happen in a real election is the Returning Officer would toss a coin, just as they would if there's a tie for the winner in an ordinary first-past-the-post election. Published results would look like this:

  • 1st count:
    • candidate A: 25 1st preferences
    • B: 16
    • C: 12
  • 2nd count: C is eliminated; 12 votes redistributed as follows:
    • A: 7
    • B: 3
    • no further preference: 2
  • result after 2nd count:
    • A: 32
    • B: 19

Would this satisfy your idea of public availability? Peter Jackson 09:59, 2 March 2012 (UTC)

Well, I don't think so, but perhaps I am wrong. For someone to run the algorithm independently, wouldn't they need all of the preference information? For example,
  • Candidate A: 25 1st preferences, 20 2nd preferences, 10 3rd preferences.
  • Candidate B: 16 1st preferences, 7 2nd preferences, 32 3rd preferences.
  • Candidate C: 12 1st preferences, 3 2nd preferences, 40 3rd preferences. Dan Nessett 18:53, 2 March 2012 (UTC)

Now to STV, which is a more complicated version of the same method. This is for when you're filling more than 1 place together: 2 or 3 seats on MC or EC. Let's take 3 as an example. Start by counting the number of valid ballots. Divide by 4. That's the quota. Anyone with at least that many votes is necessarily elected. (It's obviously impossible for too many to be so, unless all the votes are equally split among 4 candidates, in which case any election method requires coin-tossing.) If anyone is so elected, you redistribute their surplus votes, that is, the difference between what they got and the quota. (If there's more than 1, start with the one with most votes.) You do this by a proportion sum, which is where a pocket calculator comes in handy. E.g., if the quota is 5 and someone gets 7 votes, then each vote is transferred to the next preference, but marked down to count as 2/7 of a vote. If, however, the number of votes with a further preference marked is less than the surplus, you don't mark them up to compensate.

Once you've done this for all elected candidates you start eliminating the bottom ones as in AV.

An additional complication is that, as some votes disappear from the total because no further preferences are marked, you recalculate a reduced quota from the number of remaining votes (including the quota for elected candidates. Anyone reaching this reduced quota is elected, but only surpluses above the original quota are redistributed.

Published results would be similar to the above AV example, but obviously mre complicated. I hope this is all clear enough. Peter Jackson 10:16, 2 March 2012 (UTC)

Supplementary note: on the question of deciding which equal bottom candidate is eliminated, note that one common case where it makes no difference (in AV at least) is where, for example, the 2 bottom candidates have 1 vote each and the next up has 3 or more. Peter Jackson 10:18, 2 March 2012 (UTC)

From your description it appears voters specify preferences as in AV voting. However, it isn't clear from your description how these preferences are used. For example, you write, "Start by counting the number of valid ballots. Divide by 4. That's the quota. Anyone with at least that many votes is necessarily elected." Do you mean anyone with at least that many 1st preference votes is necessarily elected? Or do you mean something else?
More generally, your description, while useful in giving the "flavor" of the vote counting algorithm is more indicative than declarative. Surely, there is a precise definition of the algorithm somewhere that you can cite.
Eliding for the moment the question of whether someone with little or no experience in algorithm specification/use could implement STV with pencil and paper, it isn't clear to me from the general description you give what exactly it achieves. I was under the impression that STV eliminated the possibility of tie votes. But, it appears not. Right now, if we have a tie vote (which, by the way, has never happened), we execute a run-off election. You suggest using a random draw to break ties. We could change our procedures to use a random draw and eliminate run-off elections, so STV doesn't appear to provide any advantage in that regard. So, 1) what problem does STV solve; and 2) can you provide any evidence that this problem exists in our community? Dan Nessett 18:53, 2 March 2012 (UTC)
Little time now, so only brief/partial replies.
Your interpretation on the first point is correct.
If you want citations ask Peter. He's the expert here.
One problem is sensitivity/chaos. In an election for 3 members, a change of 1 vote can change all 3 seats in the current block voting; STV only 1. Peter Jackson 10:58, 3 March 2012 (UTC)

Also pressed for time right now, but I'll return later. Just to note, for one of the elections on which I served on the committee, we used OpenSTV software. This is available for download from www.openstv.org (and I note that they appear to now offer secure online voting, for free, which is also something we should look at). The software in question allows a choice of election methods - AV, STV, Concorde, etc. For the electorate size we have, it's perfectly possible to perform an election using pen and paper (or, more realistically, as we are geographically dispersed, an online spreadsheet). OpenSTV just makes the process quicker and simpler and reduces the capacity for human error. It is perfectly possible to get exactly the same result the old fashioned way. The software really is simple to use and does not require training - you simply enter the election method being used, the names of the candidates, and the preferences recorded for each candidate. The output in each round of voting can be copied and pasted to show the results at each stage. Anton Sweeney 14:52, 4 March 2012 (UTC)

Well, Anton, while people keep saying it would be "possible to perform an election using pen and paper" with STV, no one seems willing to specify the algorithm necessary to accomplish this. Once again, relying on software that implements an unspecified algorithm would result in a loss of transparency. Dan Nessett 18:40, 5 March 2012 (UTC)
An explanation of STV is here. An explanation of the methods available for counting (the algorithm, I suppose) is here. Anton Sweeney 00:04, 6 March 2012 (UTC)
Thanks. However, what is required is an explanation that someone can use to actually process published results; something like this where I provide the vote counting algorithm for plurality voting. Dan Nessett 00:26, 6 March 2012 (UTC)

That can't be done unless we know exactly which method of STV is being used (there are several). But in general: 1. Determine the quota. The Droop formula is the most common for doing this: quota = (valid votes cast/(number of seats +1)) + 1. (*) 2. Any candidate who has reached or exceeded the quota is declared elected. 3. If a candidate has more votes than the quota, that candidate's surplus votes are transferred to other candidates. Votes that would have gone to the winner instead go to the next preference listed on their ballot. 4. If no one new meets the quota, the candidate with the fewest votes is eliminated and that candidate's votes are transferred. 5. This process repeats until either a winner is found for every seat or there are as many seats as remaining candidates. (*) So if we've 40 voters taking part in a ballot to elect 3 candidates, the quota is (40 / (3+1)) + 1 = 11.

But really - read the WP article. It explains very clearly how STV works and how to do a count and does a better job of explaining it than I can. Download the OpenSTV software and "run" an election using sample data, then use the WP page's explanation to do it manually, and you'll arrive at the same result. An algorithm isn't necessary; a list of steps or instructions might be. Anton Sweeney 14:58, 6 March 2012 (UTC)

Thanks for making an attempt to explain the vote counting algorithm, Anton. However, I think you have demonstrated the point that I am making. The steps you give are missing important information, which someone who is not familiar with the STV vote counting algorithm would need to implement it. For example:
  1. Do voters specify preferences when they vote? If so, how does this relate to the criterion, "[a]ny candidate who has reached or exceeded the quota is declared elected"?
  2. How is a surplus vote defined?
  3. How explicitly are they "transferred"?
  4. When a candidate is eliminated, what happens to all of his/her preference votes (if STV stipulates the use of preference voting)?
I'm afraid just telling everyone to read the WP article insufficient. Our voting procedures need to be self-contained. Also, suggesting downloading the STV software and running it doesn't demonstrate how to execute the vote counting method using only pencil and paper.
So far I have been participating in this discussion as a private citizen. However, given the lack of success in anyone explaining how the STV vote counting algorithm works (in sufficient detail so a non-technical member can implement it using pencil and paper), I will now put on my MC hat and assert the following. Unless, the latter is accomplished, I will neither introduce an MC motion to use STV voting in the next election nor will I vote in favor of such a motion introduced by others. I hope that gets the attention of those supporting this voting method and motivates them to do the work necessary to accurately and precisely define the algorithm.Dan Nessett 17:51, 6 March 2012 (UTC)
1. Is this not self-evident?
2. Subtract quota from votes obtained. If the result is positive then the result is the number of surplus votes. If negative, there is no surplus.
3. That depends entirely on which STV method is being used.
4. They are transferred to the next preferred candidate in accordance with the STV method being used, or where no additional preference is stated, they are discarded.
STV is more complicated than first-past-the-post, but it is fairer and results in less 'wasted' votes. If what you're looking for is a set of clear and precise instructions for carrying out a vote count, then yes, I can provide that if I arbitrarily decide on a quota type and STV method. My explanation will not, however, be much more than an expansion on those in the WP articles, which you seem reluctant to read. I'm on holiday from Friday until Tuesday, with very limited 'net access, and therefore under time pressure until Friday, so I may not be able to furnish what you're looking for until sometime next week. Anton Sweeney 11:23, 7 March 2012 (UTC)
1. No.
2. You keep referring to "votes", but if ballots specify preferences, to which "votes" do you refer?
3. Well, which STV method are you and others proposing? Define it.
4. Again, provide a precise definition of the algorithm you and others propose to use.
In what sense is STV voting "fairer"? Fairer than what? How do you define fairness?
You write, "If what you're looking for is a set of clear and precise instructions for carrying out a vote count, then yes, I can provide that if I arbitrarily decide on a quota type and STV method." Please do so. Pick the version of STV you propose to use and explicitly define the algorithm in sufficient detail that a non-techncial member can execute it using published election results (which you are free to define) and using only pencil and paper.
In regards to my reluctance to read the WP article, I have read it. However, it should not be a requirement for its use that every member reads it. This exercise is for the benefit of Gareth, who must return a report to the MC explaining the pros and cons of the various voting methods. Up to this point in regards to STV there has been a lot of proof by emphatic assertion and not enough sound argument. Dan Nessett 17:33, 7 March 2012 (UTC)
1. Yes, voters express preferences; they put a '1' beside their preferred candidate's name, and optionally a '2' beside their second choice, a '3' beside their third choice, and so on, until they have run out of candidates or decide not to continue expressing preferences. So if candidates are Alpha, Beta, Charlie and Delta and I want Beta to win, wouldn't mind Charlie being elected, wouldn't be too put out if Alpha got in, and definitely don't want Delta to get in, my ballot might look like:

Alpha 3
Beta 1
Charlie 2
Delta

2. In any one count, votes are the total of the #1 preferences received by that candidate, plus any preferences transferred to that candidate following the distribution of a surplus after another candidate has been elected with a surplus over the quota in a previous count, plus any preferences transferred to that candidate following the elimination of other candidates in a previous count.
3. There are several different methods, as you know. I'll have a look over them and see which I think would be most suitable for CZ; in one of the 2010(?) elections, the Northern Ireland method was used. Anton Sweeney 19:20, 7 March 2012 (UTC)

[unindent]

Peter Schmitt has provided some references to material on various voting methods. These are welcome. However, do those who support the use of STV voting expect Gareth to read this material and extract from it a pros and cons analysis of STV? If so, then I think their expectations are unrealistic. It is up to those in favor of STV to make the case for it. So, we still need:

  1. An explicit description of the STV variant that its proponents are proposing. This description should be sufficient that a non-technical member can use it with pencil and paper to confirm the results of an election using published election results.
  2. A discussion of the advantages of STV voting. This discussion should define its terms, e.g., what is meant by "fair". It should also demonstrate other properties that are attributed to STV, e.g., that it is less likely to lead to tie votes than Plurality voting.
  3. A discussion of the disadvantages of STV voting. While the proponents of STV may be reluctant to do this, without such discussion, the argument in favor of STV looses credibility. Dan Nessett 22:04, 7 March 2012 (UTC)

Simple-majority voting

I think we also need to set out the reasons for and against maintaining the current system. The main point in its favour seems to be simplicity. But for me there are three main objections: firstly, candidates can be elected without an absolute majority (over 50%) of votes cast. Secondly, many votes are wasted. An illustration, using the last EC elections:

  • Author candidates: *A 13 votes, *B 13 votes, C 8 votes, D 7 votes, E 6 votes, F 6 votes (total: 53 votes)
  • Editor candidates: *W 17 votes, *X 16 votes, Y 14 votes, Z 7 votes (total: 54 votes)
  • (*) = elected. 27 or 28 voters.

For the Author contest, neither successful candidate received an absolute majority, i.e. 14 out of 27 people declined to support their candidacies. Both Editors received absolute majorities. For the Author vote, 26 out of 53 votes had an influence; the other 27 may as well not have been cast (about 51% of the total). Likewise, 21 votes for unsuccessful Editor candidates were wasted (39%). For the Authors, the majority of votes were wasted.

The third objection is that the system is likely to produce tied votes, especially when there are low numbers of voters and several strong candidates. This has been narrowly avoided up to now, in that there have been ties but only when all candidates concerned experienced the same outcome. The problem is that votes cannot be transferred to break ties, making a run-off election much more likely. John Stephenson 12:55, 3 March 2012 (UTC)

In regards to the advantages of Simple Majority voting (more precisely, Plurality voting), its simplicity supplies it with an important property - understandability. Even the least technical member should have no trouble understanding how it works.
Plurality voting has two further advantages: 1) transparency, and 2) maintenance of secret ballot privacy by making it unnecessary to publish individual ballots in order to provide transparency. The latter property is shared by the AV method and may be shared by the STV method (to determine the latter requires a precise description of the STV vote counting mechanism). Transparency means a citizen may take the published election information and independently determine who was elected.
For completeness, here is how a citizen may compute the results of a vote from published results:
  1. Order the candidates by the number of votes they received (this tally being published by the election committee).
  2. If there are n open seats. select the top n candidates in this list. All candidates receiving more votes than the nth candidate are elected. If the n+1st candidate received less votes than the nth candidate, all n candidates are elected. If the n+1st candidate received the same number of votes as the nth candidate, either 1) hold a runoff election between the candidates receiving the same number of votes as the nth candidate (the current procedure), or 2) choose from those candidates receiving the same number of votes as the nth candidate using a random draw. Dan Nessett 17:24, 3 March 2012 (UTC)
An even simpler version of the Plurality voting algorithm is:
  1. Order the candidates by the number of votes they received (this tally being published by the election committee).
  2. If there are n open seats, compare the votes cast for the nth and n+1st candidate. If the latter is less than the former, the top n candidates are elected. Otherwise, those candidates receiving more votes than the nth candidate are elected and either 1) hold a runoff election between the candidates receiving the same number of votes as the nth candidate for the remaining seats (the current procedure), or 2) for the remaining seats choose from those candidates receiving the same number of votes as the nth candidate using a random draw. Dan Nessett 18:42, 3 March 2012 (UTC)

I don't see any reason why it would be necessary to publish individual ballots if using STV. The results of each count can be published, so people can see where transfers went and how the final result was arrived at, but this does not require the ballots to be published, only the totals at each round. The OpenSTV software allows for this. Anton Sweeney 00:15, 6 March 2012 (UTC)

I don't know that publishing individual ballots would be necessary. But then again, no one has actually provided a precise description of the STV vote counting algorithm, as I provide above for plurality voting, so I don't know what results would have to be published in order for a non-technical member to independently validate an election result. Dan Nessett 00:31, 6 March 2012 (UTC)

Some references

Forum thread

From The mathematics of voting/Bibliography

  • Keith Devlin, The perplexing mathematics of presidential elections. (November 2000)
in: Devlin's Angle at the Mathematical Association of America
all in: Feature Column. Monthly Essays on Mathematical Topics (Archive at the American Mathematical Society)
Retrieved from "https://citizendium.org/wiki/index.php?title=CZ_Talk:Ombudsman/Archive_1/Election_process&oldid=837655"