42 thoughts on “A guest post on Condorcet voting”

1. Looks like word doesn’t format html well at all.

2. Word doesn’t format anything well at all. I tend to stick to text editors, and if I need formatting I type in the markup required for HTML or LaTeX myself.

3. How can you talk of Condorcet voting at a time like this? Essendon are on the verge of beating West Coast at Subiaco. Its 11.00pm and I feel like having a beer in celebration.

4. Hold that beer, Benno, you just got done.

5. I’ll covert it to pdf if you like.

   STV has some fairly obvious warts that Condorcet voting addresses, but then Condorcet voting has its own warts.

   A side issue: STV translates well to a multi-member system with a high degree of proportional representation. Near as I can tell, Condorcet voting does not (you end up with a majority-take-all system).

6. JQ, precisely because I am a Microsoft avoider, I have had to get a couple of utilities that usually do a fair job of unpicking a word file (sometimes one or the other can’t cope).

   If you want to send me the offending file, I can give it a go. Alternatively you can google for “antiword” to see one of these utilities, and maybe others of that ilk will show up.

7. Oh, if all else fails, try cutting and pasting the word material into a text editor, then putting the preformatted tags around it before saving it with an htm or html extension.

8. Well I suppose that having a beer at 11.30pm wouldn’t have been too healthy for me anyway, but next year we will be so much better. I’m still happpy because we played very well and didn’t zone out in the 4th quarter.

9. I’m curious what effect Condorcet voting would have on the electoral cycle. As a more sophisticated method of measuring peoples opinions is it possible that it could put an end to election day parties? Would results be in time for the traditional celebration/drowning of sorrows?

   Perhaps it could be hosted as a word document for download to solve the display trouble?

10. It might mean the end of the old Mackerras Pendulums, as more than the preferences between the major parties would matter. But if we continued to have a two major party system where the Coalition and the ALP combined gained 90% of lower house seats, then I don’t think it would affect the electoral cycle. A big unknown is how it would affect the political landscape, which is why I advocate that the AEC trial it after the official results on election night as a research project.

    I imagine that results would take about the same ammount of time to count, as several glances are required to count one ballot, but after that it doesn’t need to be looked at. But this is another unknown, so again a good reason to trial it. If nothing else comes of Condorcet after the dress-rehearsal, it would at least be an interesting thing to analyse.

11. That HTML version has its first few lines overwritten by blog generated stuff, at least on my software. That is, it is less readable than the usual stuff at the head of a posting – the headers usually work properly.

    I think I’ll forward Benno the HTMLised rendering I just sent JQ, for comparison. It’s probably too lengthy to justify pasting it or its text working version in here.

12. The problem I have with more advanced single-member voting systems is that they are too difficult to explain to the non-expert. Une of the few institution that does use Concordet voting is Debian, a worldwide collective of computer programmers who put together a version of Linux. But that crowd is self-selected for exceptional mathematical abilities.

    But then again, how many punters actually understand how our current system works anyway?

13. My experience (running single- and multi-member STV elections for the Green Party in the US) is that, if the voting process itself is simple–and most voters grasp the concept of ranking easily enough–, then familiarity breeds content.

    That contented familiarity, though, depends on having a system that doesn’t reward strategic voting, and that’s where, it seems to me, Condorcet voting has the edge over single-member STV.

14. While I do give a detailed example of how STV can still be subject to strategic voting, in Australian single member elections this happens rarely because of the dominance of the two parties, which means that there is rarely more than two strong candidates. So to be fair strategic voting is a big concern with FPTP but only a small concern with STV in the current political climate, this climate of course can change, which is went close to doing in Queensland with the rise of One Nation.

    Having said that though is it our STV system which artifically supports the two party duopoly, or is it something else? Maybe with a fairer voting system we will have more minor party and independent candidates winning seats, and thus encouraging more people to vote for them because they know that it is worthwhile (a different thing to not wasting a vote). So many unknowns.

    I would phrase the reason that Condorcet has the edge over STV differently, or maybe it is a different reason. In Condorcet all of the people all of the time have their full preferences recorded, everyone has a maximal impact on the outcome. Whereas in STV, the full impact of your vote depends very much on how many other people have voted in a similar way.

    So all Libertarians out there will support Condorcet on that principle.

    Previously Jonathan you mentioned that Condorcet has its own warts as well. Are you referring to Ambiguity resolutions and/or methods of counting?

15. What do you say to the criticisms of John Quiggin

    Taking this a bit further it seems likely that, whatever rule is chosen for resolving cycles, an implementable Condorcet system would be vulnerable to exploitation by strategic choices to run (or not run) particular candidates whose function would be to tip the balance in favor of some other candidate.

    Moreover, every scrutineer I’ve ever met has a clear idea of how to count an STV ballot. I just can’t say the same applies to a Condorcet ballot where you need to do a little more than just arrange ballot papers into parcels.

16. Condorcet is unmanipulable in this way. And as far as shown unmanipulable in all ways.

    John Quiggin presents absolutely no evidence to back up those assertions which are vaugely worded, much as politicians word things as demonstrated here.

    So I say that those criticisms aren’t well formed enough to be analysed. But I really hate being forced to honestly criticise someone who has been kind enough to indulge my interests. I am not saying outright that no founded criticisms of Condorcet exist, just that none have been presented to me.

    So Condorcet ballot papers are hard to count because hardly anybody has counted them before? Because nobody can immediately visualise in their head how to count them? Well French is hard to speek if you’ve never heard it before. Is there a gene for bad argument?

    I don’t think there is a gene for bad argument, so I will go out on a limb and say that it is probably the case the people who criticise Condorcet without proper arguments and evidence are closet conservatives.

17. If it makes you fell better and simulteanously protects me from attacks based upon brutal honesty, here is my latest post which is so horrible I will take it down after a while. http://australianpolitics.blogspot.com/2005/06/swipe-at-print-commentariat.html

18. Those who are attracted to the theoretical elegance of the Condorcet model are guilty of thinking too hard, a shortcoming shared by those who concocted the voting system for the Senate (the attraction of which is that every single vote is chopped into pieces and divided among various candidates, but not a fraction of any vote exhausts – beautiful, but it produces absurdities like last year’s Family First win in Victoria). Yes, it may well be that a majority of voters in Wentworth preferred Peter King to Malcolm Turnbull, but only 18 per cent actually voted for King. If I was one of the 41.8 per cent who voted for Turnbull, I wouldn’t be much interested in arguments to the effect that King deserved to win.

    It is true that strategic voting is a theoretical possibility under STV. At the recent WA election, Labor voters in the seat of Vasse who preferred the ex-Liberal independent incumbent (Bernie Masters) to the Liberal candidate who bumped him aside for preselection (Troy Buswell) would have been advised to have given Masters their primary vote, since Labor had no chance whereas Masters needed to finish ahead of Labor to get their preferences and pull ahead of the Liberal. Masters in fact ended up a mere 124 votes ahead of Labor (out of 12,247 valid votes) at the penultimate count, and then lost to the Liberal by only 209 votes.

    http://www.electionswa.com.au/_la_districts.htm

    However, I think the number of voters in Vasse who foresaw all this before polling day would have been extremely small. Masters no doubt won primary votes from Labor supporters who reasoned he had a better chance of winning, but I suspect this would have been true under any voting system – including Condorcet, even though it would remove the rationale for doing so.

    Herein lies the problem. I concur with the comments posted by Robert Merkel (who says Condorcet-style systems are “too difficult to explain to the non-expert”), and Jonathon Lundell (“if the voting process itself is simple, then familiarity breeds content”), and I query the qualifications they hasten to add, which suggest that the current system is no better understood than Condorcet would be. Condorcet stumbles at a hurdle that STV clears, in that people COULD understand STV if it were clearly explained to them and they had a minute or two to think about it. This is an important point. People can happily live in ignorance of the workings of the electoral system for most of their lives, but when an election outcome is contentious it is important that the authorities have a sporting chance of being able to explain what happened.

    Eexcept insofar as single-member electorates are inherently contentious, I don’t think there’s much wrong with the current system that can’t be fixed by simple optional preferential voting, as operates in New South Wales and Queensland. OPV allows people who know they are voting for a fringe candidate to both indicate where their sympathies lie and have a say in who gets elected, if that matters to them. It also removes from voters the onus of calibrating their opinion of every one of as many as 15 candidates. Despite the claim in Spearritt’s paper that Condorcet could also accommodate optional preferential voting, it would in fact further complicate a system that is already too abstract by half (under full preferential Condorcet voting, there would at least be a reasonably limited number of potential permutations of voting order).

    OPV is also the solution to the Senate’s woes, because it would mean that the seats that didn’t go to candidates with a full quota would go to those with the nearest thing to it on the primary vote, unless enough voters consciously directed their preferences otherwise. The Greens currently have a bill before parliament to require that voters number every box above the line rather than just one. This misses the point – voters should be able to number as many boxes above the line as they wish, and where no preference is indicated their votes should exhaust.

19. Benno- I was referring to ambiguity resolution. The proliferation of Codorcet disambiguation mechanisms might be taken as evidence that we’re not entirely comfortable with any one of them.

    I’ve heard it argued that the presence of ambiguity in Condorcet voting is a strength rather than a weakness, because it reveals an underlying ambiguity in the voters’ opinions. I don’t think “ambiguity” is quite the right word here; it’s more that different voters rank the importance of issues differently, but I think that the observation is correct.

    But when an ambiguous result arises, we’re constrained to deal with it; after all, we have only one seat to fill. While Condorcet results might *reveal* underlying ambiguity, we’re not entirely settled on how to *resolve* that ambiguity.

    (I wish WordPress offered a preview option for comments. In my earlier post, I meant dashes, not strikeout.)

20. Bludger- I agree that single-member STV is a whole lot easier to _count_ than is, say, CSSD. But it doesn’t necessarily follow that STV is easier to _understand_ than CSSD. What we mean by “understand” goes beyond the ability to perform or describe the counting process to a grasp of how the system can fail, and what it’s actually measuring.

    WRT the dominance of two-party systems, I don’t think we’ll address that with any form of single-member voting.

21. Poll Bludger,

    “but only 18 per cent actually voted for King. ” – That is the point, in single member STV the only vote that you know will count is your first preference.

    If you were one of the 58.2% that didn’t vote for Turnbull then you would be very much interested in system that fully noted your preferences. Or are you advocating a return to FPTP?

    Your example of strategic voting in WA provides nothing for your case.

    Your paragraph about the dificulty of explanation is a good point. But it is a very easy hurdle to overcome. So if that is the only problem then I ask “What’s the problem?”

    Your second last paragraph has a lot of merit, I support the introduction of optional preferencing for STV. You are right that full preferential would be much better for Condorcet and I would say that Condorcet with optional preferential is better than Condorcet with compulsory preferential which is better than STV with optional preferential. You are wrong to say that Optional preferential STV is just as good as Condorcet though.

    I completely agree with your last para.

22. Just clarifying something there.

    full preferential Condorcet would be better from a practical voting point rather than the best theoretical method, which is provided by optional preferential condorcet.

23. Johnathan,

    “The proliferation of Codorcet disambiguation mechanisms might be taken as evidence that we’re not entirely comfortable with any one of them.”

    That is a fair point, but I would take it as evidence that there are many ways to skin a cat, I wouldn’t necessarily make a further statement to that. But if we have a Condorcet system, the only thing that seperates each is how to resolve ambiguities, so we need to pick the best way.

    In all cases you take the Schwartz set, which I have explained. These members of the Schwartz set are all equally our Condorcet ‘winners’ if no one winner exists. So a method of resolution will choose one of these as our winner.

    I hope that I have explained CSSD sufficiently, if not I can explain it again. If you have a look at the different methods and particularly the link to http://www.electionmethods.org who advocate CSSD it should be pretty clear if you spend some time to mull it over.

    In other words the American based Condorcet community all think that CSSD is the way to go.

    ps.

    I would have to say that anyone clutching to the problem of the explanation of Condorcet voting is a closet conservative bureaucrat with no imagination who is clutching at straws. In all of your self righteous holier-than-thou text Poll Bludger, that is the only substantiated quibble you came up with.

    Now to go out on another limb. Perhaps the reason you and Antony Green are so disdainful of Condorcet without having proper arguments is that you both have a vested interest in the maintanence of the status quo. Either that or you have little imagination.

    Alan, change your url in the comment section, a typo sends people to bible bashers heaven.

24. The proposed method reminds me of the finals method that they used to use for the AFL. It might have been right, but nobody could explain it clearly.

25. Benno, my Vasse example was not supposed to support my case – I was in fact playing my own devil’s advocate. If I wanted to illustrate the undoubted theroetical advantage that Condorcet has in removing incentives for strategic voting, that’s an example I would use.

    If I was one of the 58.2 per cent who didn’t vote for Turnbull I would still look at the primary vote figures – Turnbull 41.8 per cent, Labor 26.3 per cent, Peter King 18.0 per cent – and accept that whatever my personal preference, Turnbull won fair and square. If I was a Turnbull supporter and King won, I would not be so understanding.

    I don’t advocate a return to FPTP, but it does have points in its favour (as does Condorcet – it doesn’t pay to be too dogmatic about these matters). If you don’t arrive at or near first place on the primary vote, I don’t think you should be in the hunt, and I think most voters would have a hard time understanding arguments to the contrary.

    I support OPV over full STV for a similar reason – it will deliver victory to the leader on the primary vote more often, while still assuaging Ralph Nader-style spoiling effects. I grant you that it lacks Condorcet’s conceptual elegance because it involves a compromise between two principles, but I think both principles are important and that both Condorcet and full STV give one of them too little weight.

    I can assure you I have no interest in the status quo, vested or otherwise -in fact, as an advocate of OPV, I don’t even support it. I also doubt that Antony Green would be out of job if Condorcet were introduced. In fact I think he’d have a field day.

26. “and accept that whatever my personal preference, Turnbull won fair and square.” That is a throw back to the dark ages of FPTP, whatever you say. The point of IRV and even more so in Condorcet is that all preferences are counted to determine the candidate which a majority prefers over all others. So the point is that we aren’t just blindly looking at primary votes, otherwise we might as well have a FPTP system, which Turnbull would have one as well.

    “If you don’t arrive at or near first place on the primary vote, I don’t think you should be in the hunt, and I think most voters would have a hard time understanding arguments to the contrary.” – A worthy point perhaps, but it ignores the possibilty of clone candidates which would very much exploit a primary vote cut off of say 5%. My other changes to candidate registration ensure a smaller number of candidates with much more electorate approval which pretty much negates this argument, all without the need for arbitary cut-offs or arbitary voting methods such as STV with or without compulsory preferencing. Primary votes aren’t and shouldn’t be the be all and end all, which is the point of any preferential system, as not everyone will get their first preference up in a single member election. Hence the need to at least have the most say in the winner. If you want to continue your belief in the holiness of first preferences then I suggest proportional representation for the lower house, but not a bastardisation of single member democracy.

    “I grant you that it lacks Condorcet’s conceptual elegance” – You are mistaken in thinking that I am attracted to Condorcet for conceptual elegance. I am attracted to it for it’s mathematic simplicity and democratic superiority. And yes it is mathematically simple.

    “I can assure you I have no interest in the status quo, vested or otherwise -in fact, as an advocate of OPV, I don’t even support it. I also doubt that Antony Green would be out of job if Condorcet were introduced. In fact I think he’d have a field day.”

    That is all true, I didn’t think Green would suffer a loss of income, just that he had a vested interest in the status quo. What form that would take I have no idea, just wondering why the psephological set are unreasonably anti-Condorcet. I threw you in as ‘possibly having a vested interest’ to bait you, which didn’t work very well.

    But aside from all of that Mr Bludger. Suppose that you are largely right (not to hard to suppose), but that you also thought kindly towards misguided souls such as myself, which no doubt you do. Now I am interested in a Condorcet count done after the 2007 federal election results, the count would be done in seats where a winner didn’t exist on first preferences. At the last election 89 seats were decided after the primary count. Anyway how do I convince someone preferably the AEC to do this project? Also let’s consider the possibility that being bureaucrats they have no imagination. So what to do?

27. So Condorcet ballot papers are hard to count because hardly anybody has counted them before? Because nobody can immediately visualise in their head how to count them? Well French is hard to speek if you’ve never heard it before. Is there a gene for bad argument?

    I don’t think there is a gene for bad argument, so I will go out on a limb and say that it is probably the case the people who criticise Condorcet without proper arguments and evidence are closet conservatives.

    I am talking about the physical act of counting the papers, not previous experience. I would regard your acerbic answer as merely par fr the course. However on this occasion, as with the famous invention of inexhaustible transferable voting, you’ve simply missed the point.

    An STV tally is easy to double check because as the tally progresses the ballots get sorted more and more finely. A Condorcet tally must completely re-sort the ballots for each pair-wise comparison. There is no reason to expect each re-sort to retain anything in common with the previous re-sort. You end up moving to Ohio and not having a verifable paper trail, or you re-sort and re-count every time there’s a query. The problem flows from the nature of the tally, not previous experience.

28. Well you obvously don’t realise that it’s not that hard to count the ballots, I have no idea what you mean by all of this “A Condorcet tally must completely re-sort the ballots for each pair-wise comparison. There is no reason to expect each re-sort to retain anything in common with the previous re-sort.”

    It is utter nonsensical, confusing, Howard esque Bulls#it. See Bartlett’s explanation of Howard lingo.

    One ballot is looked at and the results of which are added to the one tally point of the matrix, which is best viewed in the word document. There is no ‘re-sorting’ and there most certainly is a paper trail. Come back when you have some good arguments.

    While I completely missed the point with my inexhaustible transferrable vote, you have missed the point on this occasion. Stop peddling scare tactics about ‘how impossible it is to count a condorcet election’ and ‘we won’t have a paper trail so we will have a system open to rigging like in the US’. It is nonsense goddamit! And dirty filthy lies to boot.

29. You don’t do your own cause much good by accusing every second poster of lying. You still need to look at the paper ballots, even if you enter every ballot into a Word document. If a discrepancy shows up (and it’s not immediately obvious how a Condorcet discrepancy can be detected) you need to look at ballot papers, not entries in a Word document.

    Every time you do a pairwise count you need to re-sort the papers. Either that or you have to individually mark every ballot to test that the tally entry is accurate. If you can’t look at your Word tally and say that a particular vote came from a particular paper, and put your hand on that paper, then you’re back in Ohio. It’s unsurprising how few Condorcet advocates there are among those of us who’ve actually counted an election.

30. You’ve missunderstood the phrase ‘word tally’, I didn’t mean to put it into a word document and count it that way, I was just referring to the word version of my article in which I have a matrix, the matrix where votes are tallied. It read badly so my bad.

    “You don’t do your own cause much good by accusing every second poster of lying.” – I don’t suppose I do, but then you don’t do your cause much good by lying, oh actually you do, damn. Maybe though your argument rather than being lies was just very hard to understand. It is the peddling of small matters as big problems for condorcet that I regard as lies, or at least miss-information.

    Now, your second para is confusing, I don’t understand what the problem is. I will slowly and methodically rationally disect it for the benefit of all.

    “Every time you do a pairwise count you need to re-sort the papers. ”

    No is the short answer. The way it is counted does not mean that:
    You first count ballots for A vs B, then recount every single ballot to determine B vs C, then recount for C vs D etc…

    It doesn’t operate like that. First you look at ballot paper 1, you see that the voter has ranked candidates as such: a,b,c,d,e

    In some sort of matrix, which takes the form as appears in the word version of my article (the html and post version don’t format as well), you count all of these preferences at the same time. in the A vs B column you add one vote to A and zero votes for B, in the A vs C column you add 1 vote to A and zero votes for C. etc… time consuming? possibly. If the AEC do the count then we will see.

    So one ballot takes a long time to fully count, but after that you don’t need to count it further. Whereas one ballot for STV takes much less time to count, but chances are that you will need to recount it at some point it the tally, to look at second preference, third preference, etc.. until somewhere on the ballot is recorded one of the ‘two candidate prefered’ candidates.

    To speed up the process and reduce election night hangovers the first count could be done by computer, keep in mind that not all ballots need to be done on the same computer. different polling booths can tally ballots in different locations without any need for communication, unlike STV where a first count needs to be done and universally tallied before it is decided which candidate to redistribute votes for.

    At the end of each computer in each polling booths tally it is simple to add the different matrixes together. Once that is done then the condorcet winner is decided. After the first count (which would be incredibly quick) a hand count can go ahead for the official results. But all things considered I don’t see why a hand count for Condorcet would take longer than a hand count for STV. Unless you had a rediculous number of candiates, but my further suggestions negate the possibility of too many candidates contesting.

    “It’s unsurprising how few Condorcet advocates there are among those of us who’ve actually counted an election.” That is the most arrogant thing I have read in a long time and I do read my own writing. Not to mention misleading as there are hardly any condorcet advocates in the general population, so statistically speaking it would be very unlikely to come across any condorcet advocate, even an election scrutineer/counter.

    That is an example of a lie.

31. Who’s next?

32. Benno, dude, this is great.

    But if you ever want to get it implemented you need to make it easy for the public (aka those dumb bastards) to understand.

    As soon as I get back from this place (China) I will begin work on a slideshow of pretty backgrounds and cartoon characters to inform (bias) the public.

33. Yay, rejoicing in the streets. Will it be based on THE MAN comics?

34. I’ve been doing computer-assisted counts of multi-member STV elections for a while, now, and I don’t see why the same general approach couldn’t be used for a Condorcet count. The process goes like this:

    1. Ballots are transcribed into a text file, one ballot per line. We do this with a ballot reader, a data entry person, and one or more observers. Ideally, we do it twice, independently, and compare the results. The ballots are numbered during the first data entry, so that the file(s) and paper ballots can be correlated later.

    2. The result is counted by a (relatively) simple Perl script. The script is published on the web, along with its MD5 checksum, so that it can be independently verified. (I chose Perl because it’s fairly universal, and can be run on most commonly available platforms.)

    3. The ballot file itself is also published, so that it can be independently counted, with the published script or with any independent program.

    So the ballot file can be verified from the paper ballots, and the ballot file can be independently counted by anyone who cares to.

35. Always remember, “To err is human, to really foul things up requires a computer.” (The split infinitive was present in the original, as told to me anyway.)

    I will have more later, but I do not know if I will ever have the time; procrastinate now.

36. I doubt that the original said “foul” (and what’s wrong with splitting infinitives, anyway?).

    At any rate, that’s why we want to restrict the use of computers to a role in the process that’s transparent and verifiable.

37. It’s really difficult in a language like English to argue against splitting infinitives, mostly because many verbs already get split up once. The verb with the split infinitive in P. M. Lawrence’s quote was ‘to foul up’, ‘to foul’ on its own has a different meaning, and something like *’you gotta be careful not to foul up it’ is obviously horribly ungrammatical to begin with.

38. I believe the term is a-grammatical

39. Benno, you said “Yay, rejoicing in the streets. Will it be based on THE MAN comics?”

    Yes.

40. I need more people to say that my condorcet proposal sticks it to the man.

41. Your condorcet proposal sticks it to the man.

42. That was sweet. BTW I am THE MAN. So THE MANs proposal sticks it to THE MAN. Here is evidence for my themanlyness.

    

Comments are closed.