Wednesday, October 24, 2012

Getting Ginninderraed: Another For the Hare-Clark Textbooks

 [UPDATE 25 Oct, 11:30 pm At the time this piece was written, there was still some realistic doubt in my notoriously call-shy mind about whether the event described herein would happen, although it was looking very likely.  It now appears a certainty; I've even taken down the question mark.]

Hare-Clark junkies (we're a rare and tragic breed) like to talk about classic election counts as if they were great sporting fixtures, rare vintages of wine, exotic and peculiar animals, model trains  ... you get the drift.  That is, when we can even find any others of our own pathetically small order to share our monastic secrets with.

The Hare-Clark system, on the surface one of the simpler models of quota-proportional multi-candidate election without party tickets (yes, that sounds really simple, doesn't it?), has thrown up some real beauties in Tasmania.  I don't think anyone following it would forget Kim Booth's save in Bass in 2006, when he defied expectations of most observers to scrape home from well behind by 136 votes off massive Labor leakages prompting now-Greens-MHA Cassy O'Connor to declare "Bless the intricate beauty of Hare Clark!".  But that one was only a two-horse race and nowhere near as odd as the mess in Denison 2010 in which two Liberals, a Green and Andrew Wilkie contested a four-candidate race - eventually won by Liberal Elise Archer after she jumped over a ticketmate just before she would have been cut out, and then got just enough preferences from the Greens ahead of Wilkie to defeat the latter. 

Now, in the ACT seat of Ginninderra, a new chapter in the history of Hare-Clark strangeness could be being written- one that should it happen the way it now looks like it should, will be used as an example for decades of the need to be so terribly cautious when trying to predict Hare-Clark results from raw party totals or electorate polls.  Even if it doesn't happen, the near miss will be quite instructive.




I have often said that the Hare-Clark system is about candidates rather than just party totals, and that for this reason it is possible for a party that is fairly close to its first quota of votes to be beaten by a party that is seemingly much further from its second or third.

This is because of a few aspects of the way that Hare-Clark works.  At the start of a Hare-Clark count, all those candidates with more votes than the required quota are declared elected.  Their vote total is reduced to a quota, and their papers are then passed on to other candidates at reduced value.  Sometimes this puts more candidates over a quota, in which case they too have a surplus to distribute, based on votes from the count that put them over.  But sooner or later (and sometimes straight away on primary votes) a situation is reached in which no candidate not already elected has a quota.

In these cases, what happens next is that the candidate running last is cut out and their preferences distributed, then the candidate running last after that, and so on until someone hits quota and has a small surplus to distribute.  Rinse and repeat until you elect enough candidates, or exclude enough that you know those left over are the winners.  You can find far more precise descriptions of it on the web if you need to; for instance, here's the ABC's.

This exclusion from the bottom up of all those cluttering the ballot paper with tiny vote totals isn't a big deal for the serious contenders early in the count.  But exclusion from the bottom up becomes a serious matter when there are not many candidates left.  If you can't cross a quota, you need to at least be sure you're never last when the music stops.

In the ACT electorate of Ginninderra, we currently have a situation in which Labor has 2.41 quotas, the Liberals 2.03, the Greens 0.60, the Motorist Party 0.44 and tiddlers the rest.  If this was a Senate-system election for five seats, the #1 candidates for Labor and the Liberals would be elected on primary votes, their surpluses would then elect the #2 candidates for each party, and all the Green candidate would need to do would be to stay ahead of all others as candidates were cut out from the bottom up, and they would win.

But instead of one Labor candidate having all the votes and the rest having hardly any, what has happened in Ginninderra, because of the lack of a Labor ticket order plus the use of the Robson Rotation system, is that the votes are spread around between the Labor candidates with only one all that close to a quota.  A provisional distribution based on 52% of the votes cast has now suggested that at a certain stage of the count, the remaining Green candidate, Meredith Hunter, would have about 0.76 of her quota.  Labor's total would have increased compared to the Greens thanks to the preferences of the Motorists Party especially, but they would have 2.69 quotas.

(The rest of the votes would be two quotas sitting with the two elected Liberals and over half a quota wasted as exhaust by silly people who did not fill out all the boxes.  Always fill out all the boxes if you can do it without making a mistake. There's always someone you can enjoy putting last.)

At .76 plays 2.69 it sounds like the Greens are nearer their first quota than Labor is to their third.  But the catch is that the Labor votes consist of one candidate with quota, and two others who are short of a quota but each have more than the lone Green.  Therefore at this stage Hunter is the last candidate and must receive the dreaded word EXCLUDED while Bourke and Berry for Labor both get home.

The 52% of the votes that this modelling is based on are actually fairly representative of the total, but the total primary vote for Labor to this point is actually slightly better than in the 52% sample, and that for the Greens slightly worse.  So while the current projected margin of this rare Hare-Clark event, 372 votes, may not sound like much, it's more likely than not to just get bigger.  Hunter's chances of holding her seat depend on the preference flows of the rest of the count being quite a bit different to those we have seen so far, or on any dribs and drabs of remaining postals including the electoral might of a secret Canberra forest defender convention*, or something else like that.  It's close enough she might still stave off this calamity somehow, but it isn't looking good.

Why didn't people spot this danger on the night?  Well, I did, only then I just forgot about it.  I'd been warning of the possibility of such effects in ACT seats based on the none-too-flashy Patterson Research poll, like some superstitious ancient telling all the chilluns there were Serpents in the Water (even though nobody ever saw one that was more than 14 votes long) and a poster had what turned out to be a lovely prophetic response here : "Well, that clears that one up. Talk about the fog of war."  I glimpsed the monster from a distance through the fog right here on the night, didn't think it looked too likely on the figures at that stage (which weren't as good for Labor as right now) then got distracted by some nonsense scenario about the Motorist Party being in the fight, and upon disposing of that one forgot about the other and wrongly concluded there was nothing to see here, another plain vanilla 2-2-1.

Then came this, which I checked out at manic speed and suddenly realised that it was seriously on.    As far as I know pollbludger member johncanb was the source of the first online knowledge that Labor had not just a remote but a serious chance.  The first published comment I've seen came from a poster raising it on Antony Green's blog on Sunday, to which the sorely sleep-deprived commentator at first uttered the fateful words "It seems very unlikely" - but it turns out this earlier post came from johncanb's son after johncanb raised the possibility.

I have called this scenario all kinds of things when talking about it in the past.  The two-candidate trick.  The candidate effect.  The lone loser problem.  That Strange Thing that can Happen in Hare Clark (I mention it and people look for ways out of the room).  It's a wonder that I haven't used my chess vocab and just called it the Big Swindle.  But if Berry wins, from now it has a name.  The Ginninderra.  And any candidate who loses in this manner shall be said to have suffered the ultimate indignity, of Getting Ginninderraed.  Of all electoral fates, there can be not much worse than losing your seat in such a way when almost everybody thinks that you are home.  

(* I haven't researched it systematically but I believe the idea of a "backpacker vote" that causes Greens to pick up in late counting is not consistently supported by the evidence.)

Update March 2014: Brenton Best (ALP) in the seat of Braddon at the 2014 Tasmanian state election is the latest Ginninderra victim.  In this seat the Liberals polled 58.76% (3.53 quotas) to Labor's 23.24% (1.39 quotas).  Leakage from surpluses dropped the Liberals back to 3.42 to Labor's 1.44 quotas, and towards the end of the cutup preferences had so favoured Labor that Labor finished with 1.77 quotas compared to the Liberals' 3.70. Nearly half a quota had exhausted.  All else being equal this would have been a win for Labor since they were 697 votes closer to two than the Liberals were to four.  But unluckily for Best, the two Liberal candidates, Jaensch and Rylah, finished with .89 and .81 quotas respectively.  This meant that Best actually lost to Rylah by 451 votes.

Fortunately this time there was a good level of awareness in commentary on this seat that the party totals were not where the action was.


4 comments:

  1. A nice description Kevin. I worked it out (at about 3 am Sunday morning) because I have a system of 4 interlinked spreadsheets which applies the preference distribution from the Interim Distribution of electronic votes to the latest primary votes posted. As I looked through the recalculated spreadsheet I fully expected to see that Hunter had knocked out Berry as that was what everyone was predicting. So I was surprised to see Berry ahead of Hunter in the new spreadsheet at the point of elimination. And Berry was a lot of votes ahead of Hunter, so I knew unless there was a radical change in the distribution pattern in the total vote, then Berry was certain to be elected over Hunter.
    In the day and a half after 3 am Sunday morning, life got in the way of publishing online a detailed analysis, so the credit for the first detailed analysis published online goes to my son's post on Facebook on Sunday morning. He followed this up with his question Sunday afternoon to the wonderful Antony Green, asking if 3 Labor in Ginninderra was a possibility. We will treasure Antony's reply "It seems very unlikely" , because he is such a guru on psephology and is so rarely wrong. We are very lucky in Australia to have a journalist who is such an expert on elections. (Or is it a psephologist who is a journalist!) With you leaving the Tasmanian Times Kevin, we're now down to one who does both roles.
    The story of how Antony was nudged to reconsider his Sunday afternoon conclusion is a story for another day.
    JohnCanb

    ReplyDelete
  2. Not happy with the Data entry model of Hare Clarke .There is nothing more satisfying that actually throwing paper to achieve a result and watching the pollies squirm with incomprehension. Brenton was undone by the dev/burnie partisanship. he got his start from 1's from dev but the good folks from up the rest of the coast consistently put him at 26 and consequently never collected from the subsequent counts. the best "Ginninderra" was another Braddon count. Oldaker's surplus going away from Greg Peart to Michael Weldon, It brought on the Robson rotation.
    PhilD Lapoinya

    ReplyDelete
  3. Revisiting this old theory post ahead of the 2024 ACT election, looking at Brindabella. Liberals have 1 clear favourite of 5, and Labor have 2, in what's likely to be a competition for a 3rd seat each along with Greens and maybe independents. Still can't quite tell who's better positioned to win on this basis.

    Is it fair to say the general strategy is to evenly split votes between the precisely the number of candidates you could win on a good day - no more, no less?

    ReplyDelete
  4. Yes the best condition for the trick is to have one's vote as evenly split as possible between the number of seats one is trying to win, at least after any minor candidates for the same party are excluded and their preferences redistributed. But it is also possible sometimes that a party pulls it off by having one candidate elected easily then a very even split for the next two. It only works for parties that are seriously in the running for at least two seats, and usually this requires at least a quota and a half.

    ReplyDelete

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.