Sunday, October 25, 2020

ACT 2020 Final Results Review: How Did The Greens Win Six Seats?

The ACT election is over and the Labor-Green government has been returned.  Predictions that the Liberals would get even remotely close - based in many cases on unsound analysis - have been squelched, with the Liberal Party dropping to nine seats out of 25, the party's second-lowest seat share since Hare-Clark was adopted for the 1995 election.  

In an election that saw relatively minor vote swings (2.9% against the Liberals and 3.2% to the Greens) the most striking result was the Greens' spectacular seat haul, taking two seats apiece from the major parties to go from two to six seats out of 25.  They thereby won 24% of the seats off 13.5% of the vote, a feat that requires some explanation.  This is, by a very small margin in percentage terms, their highest seat share in ACT history.  By comparison in 2010 the party won just five seats with 21.6% of the primary vote in Tasmania's 25-seat Hare-Clark system, which also has five divisions with five seats in each.

The following table shows the quota results for the parties and the seat results in each seat, and also the total percentage votes and percentage of seats.

Every seat was won by the party/parties with the largest remainders over full quotas, except for the Greens defeating the Liberals in Kurrajong (of which more later).  The Greens were short of quota in four seats, and could have lost in those seats had a major party had a higher remainder, a better preference flow or a much better split of votes between candidates, but in no cases did that occur to the extent that they were beaten.  Furthermore, while Others collectively outpolled the Greens in three divisions and nearly in a fourth, preference flows between the various minor parties and independents were, as usual, weak, and no specific fourth party or independent was a major threat in the preference distributions.  While all the "big three" benefited from this in getting a greater seat share than their primary vote share, the biggest benefit went to the Greens.

The Belco Party in Ginninderra suffered from a catastrophically high 45% leakage rate when its candidates were excluded, suggesting that its 9.4% party vote contained a large number of personal votes for its individual candidates rather than for the party as a whole.   On the exclusion of the final Belco candidate, Bill Stefaniak, 46.8% of his votes exhausted.  Given that the Liberals got 64.5% of the rest of his preferences, the decision by the Belco Party to run five candidates appears to have damaged the Liberals severely, and they only just escaped losing a seat as a result.

The other notable fourth party result was by independent Fiona Carrick in Murrumbidgee.  Carrick polled 7% and briefly loomed as a possible threat to the Greens on election night, but had nowhere much to get preferences from and was slightly outperformed by the Greens on preferences anyway.  She ultimately finished sixth, 3076 votes behind the Greens.  

The Greens' nearest escape was in Brindabella.  In this division the Greens experienced poor preference flows (especially from the Shooters, Fishers and Farmers) and very high leakage rates.  They were also at risk because of the relatively even split of votes between candidates from the major parties.  At the point with seven candidates remaining, the Greens were on 0.71 quotas, Labor on 2.59 and the Liberals on 2.52, but the individual candidate votes looked like this:

Burch (ALP) 9088
Lawder (Liberal) 8364
Gentleman (ALP) 8305
Parton (Lib) 8282
Wall (Lib) 6611
Davis (Green) 6583
Werner-Gibbings (ALP) 6501

This seat was being wrongly given to Labor by the ABC for a while on election night because it had not been noticed that Taimus Werner-Gibbings was only 36 votes ahead in the provisional distribution.  It took until Wednesday for the Greens to get the lead.  Despite being ahead of both major parties on quota remainders, the Greens were behind the Liberals on individual candidate tallies, and ended up only 82 votes (.009 quotas) ahead of Labor.  A more even vote split between the three Labor candidates would have defeated Davis and made it three elections in a row that the Ginninderra Effect (warning: extremely wonky link) had cost the Greens a seat, but Labor had only two incumbents from 2016 in this seat and therefore couldn't quite accomplish it.   Finally, re this seat, a note that I have seen Davis' final victory margin over Andrew Wall given incorrectly as 520 votes; it was actually 1654 after the throwing of ALP surpluses.  


Hare-Clark systems are often trumpeted by advocates of proportional representation, but using seats with only five members in each weakens the system in that sense, as do candidate effects.  There is nothing disproportionate about the fourth parties and independents not winning any seats, since none of them polled more than 2% ACT-wide in their own right or attracted outstanding preference flows.  Based on primary vote support, 11-10-4 would be the most proportionate result, but our system also uses preferences, and the Greens' performance on preferences was pretty good.  But there's more to it than that, because the system tends to advantage parties that consistently poll most of a quota across divisions.  It was in this way that the Greens won a sixth of the state seats off little more than a tenth of the state primary votes at the 2019 half-Senate election.

As well as this, the Greens were lucky that the split of their votes between divisions enabled them to win two close contests.  This isn't the first time this has happened; they won 4/17 seats (23.5%) off 15.6% of the vote in 2008 after overtaking the Liberals on preferences from slightly behind in the then-extant seven-seat division of Molonglo.  In Tasmania, the Greens usually don't outperform their vote share because their vote is so different between electorates that a low vote share usually results in some losses, while a high vote share results in them getting 1.5 quotas but losing somewhere or other.  In the ACT, the Greens have also had some elections where they were very unlucky with the conversion of votes to seat share, often because of candidate effects.

The Greens' 6/25 seat result off only 13.5% of the primary vote can be explained by a combination of good performance on preferences (ie some of their natural vote was hidden among similarly minded left-wing micro-parties), being helped by the Territory being divided into five-seat divisions, and luck.  

Kurrajong Result

The Greens' Kurrajong result, in which they won two seats (an Australian first for a five-seat division at state or territory level) was remarkable.  Kurrajong is the most left-wing part of Canberra and there had been some informed speculation (see first comment) the Greens might win two there, but it didn't seem too likely to me even after a uComms poll of the ACT in early August showed a modest swing to the party.  What actually happened was a major swing not just to the Greens but also to generally like-minded micro-parties, such that the Greens got 23% and there was still 5% left over for the Canberra Progressives.  The changed mix or at least the changing behaviour of micro-parties, with some of the swing hidden in preferences, masked how much this electorate had said no thanks to the Canberra Liberals.  The Liberals' campaign in this seat was a trainwreck, losing one candidate to internal disagreements and another to media exposure of homophobic and other offensive views, then having it turn out on the campaign trail that the third had been the subject of false embellishing claims about his military record in Chinese media.  It's also likely that Kurrajong was the least receptive seat to the Liberals' campaign style and to the general problem of the Canberra Liberals being right-wing in a left-wing capital.

As a result of preferences, the Greens were able to turn a 4.59% deficit on primary votes into a 0.80% victory margin for Rebecca Vassarotti over Candice Burch.  This table shows net gains and losses from  various vote sources in the process:

The "Ginninderra Benefit" is the advantage to Burch compared to the net party totals as a result of the Liberals having a second candidate under quota at that point of the count.  Overall, this mostly cancelled out the fact that the Liberals lost more votes to leakage than the Greens did (I've heard the Greens deliberately took focus off leader Shane Rattenbury to try to reduce leakage issues).  The Greens' win was therefore built on their vastly better performance on preferences from other candidates, and here they owed almost as much to Labor voters (on whose preferences they beat the Liberals 2092-808) as to the various micro-parties (2543-1159).  The Liberals only needed to gain 3.0% of the total vote from 16.1% available as potential preferences to win a second seat (given exhaust) but were unable to do it.  The impact of the Labor surplus was increased by the ACT's special surplus provisions, but the Greens would have won anyway.  Overall it was a very anti-Liberal vote in Kurrajong that left the Liberals unable to assemble enough preferences for two seats.


The matter of exhaust in ACT elections is important because the ACT are national leaders in reducing informal voting.  At this election only 1.2% voted informally, which can be attributed partly to the great increase in electronic voting but also to the ACT's generous savings provisions, which allow any vote that includes a single 1 to count, irrespective of subsequent errors.  The ACT provisions are, however, sometimes criticised for increasing exhaust.  To some degree, this criticism is misplaced because the ACT's preference distributions continue to throw votes after all seats are well beyond mathematical doubt, which creates spurious exhaust. However the ACT distributions also allocate a lot of vote-values to "loss due to fraction" that I would classify as exhaust.  The following compares the official exhaust rates for this year's counts with what I consider to be the effective exhaust rate.

The average effective exhaust rate is slightly below my estimate for 2016, and this does not mean 6.5% of voters' votes exhausted at full value, either.   The total includes votes that exhausted at a partial value after helping elect one or more candidates.  Exhaust was highest in Ginninderra, and the rather high exhaust in that division had more to do with the Belco Party running five candidates (meaning that every vote that was 1-5 Belco with no other squares numbered exhausted) than with the ACT savings provisions. 

This rate of exhaust is a small price to pay for a spectacularly low informal voting rate (though the electronic voting side of things does come with privacy and security issues).  I not only agree with the comments of re-elected Chief Minister Andrew Barr who trumpeted the low informal rate as evidence that the ACT makes it easy to vote compared to the constant attempts at voter suppression in the USA, but I also think that most other states and territories, and the federal government, have things to learn here.  Particularly, the current federal, Queensland, Victorian and NT lower house systems show contempt for voters by making no attempt at useful savings provisions.

(NB A reader has alerted me to the effective exhaust rate in the ACT being avoidable and partly due to its use of the last-bundle method for surpluses.  Because votes cannot increase in value above 1, in some cases it is not possible to fill a surplus with non-exhausting votes from the last bundle, and the ACT treats the difference as loss due to fraction.  If all a candidate's votes were thrown again using the Weighted Gregory method or similar, which can be easily done for a computerised count, this problem wouldn't arise.)

An Exportable Arrangement?

Labor minority governments supported by the Greens have been very successful in the ACT.  This arrangement has now been re-elected by voters three times, and will be 16 years old at the 2024 election.  Cue feelpieces arguing that this model will work well in other states or nationwide.  But so far, it hasn't - nowhere else in Australia (excepting the odd case of SA following Kris Hanna's temporary defection to the party) has a government dependent on the Greens in a lower house survived the next election, with three such governments losing after one term in Tasmania and one federally.  It probably will work in the future somewhere else someday (perhaps Victoria is the next most likely candidate) but I really don't believe the ACT is normal enough to serve as a useful model. 

I don't believe this for three reasons: (i) the ACT is a very left-wing electorate (ii) the ACT has high political literacy levels and (iii) the ACT has relatively few resource vs environment lobby type issues (such as forestry and mining controversies).  Without at least some of these factors on its side it's hard to see that the government would have survived the 2012 election (which it won by one seat).  

When will the Canberra Liberals ever win again?  It's a few decades since any state or federal government in Australia won re-election after more than twelve years in office, yet the ACT has comfortably re-elected a government that has been there for nineteen.  However I believe that the leftward skew of the ACT, and the relatively small stakes of territory politics compared to state politics, mean that long incumbency matters less.  Furthermore, the Liberals' result at this election will leave them needing to gain seats in four of the five electorates next time, or to rely on a coalition partner doing some of the work for them (that didn't turn out well this time).  Kurrajong requires only a very small swing to win back, but the rest require swings of several percent at the least, and my estimate of the uniform swing the Liberals would need to win outright is around 11%.  Labor would have to govern quite badly for that to be a one-election project.

A word re polling accuracy: a uComms poll released in early August had Labor at 37.6% (actual 37.8), the Liberals at 38.2 (33.8), the Greens at 14.6 (13.5) and Others at 9.3 (14.9).  While only accurate on some of the vote shares, the poll made correct findings of a clear hung parliament and a swing to the Greens, and was only incorrect in finding that swing to be coming overall from Others rather than the Liberals.  Given the time between the poll and the election, such "errors" could have been largely caused by poor Liberal campaigning.  A Clubs ACT survey was more accurate as concerns the size of the Others vote, but much less accurate concerning the combined vote for the governing parties, failing to pick up the swing to the Greens.

That, unless I think of anything else, closes my coverage of the ACT election for 2020.  Thanks to readers for the very high level of interest in this election.  Election buffs have one week now to recharge all our batteries for Queensland!


  1. Love the analysis! Bring on Queensland.

  2. Thanks for your excellent coverage. It's interesting how solid the Greens vote was across all the electorates. Obviously the Greens' strongest area is Kurrajong but it's impressive they also won seats in outer-suburban Brindabella and Yerrabi. This isn't like the Nationals in other states/federally where they're do well in a small share of electorates but do badly (or don't run at all) in most. In that sense it's quite an unusual coalition government for Australia.

    Most of the pundits overestimated the Liberals prospects as they did in 2016. I think the result in Kurrajong was pretty foreseeable - they were already well-below 2 quotas last time and inner Canberra is clearly trending leftward. Somewhat similar story in Brindabella which an ABC article claimed was a "Liberal stronghold" where the Liberals were "all but assured" a third seat despite only narrowly winning it in 2016.

    People often point to the relatively high number of public servants in the ACT to explain the left's dominance, but I think it actually has more to do with the population being almost entirely urban. Leftwing parties wins the cities in most elections e.g. if Tasmania was just Hobart, Labor would consistently form government there and Labor-Greens coalitions could probably work as well as they do in the ACT.

    1. This comment has been removed by the author.

  3. Thanks for the link to my speculation.

    Yerrabi was the biggest surprise for me. The Greens campaign was quite muted there and it was their worst result, yet a clear win on the night.

    What happened there? Is it fair to say that the ALP to Lib swing allowed the Greens to "come up the middle"? There was a short lived flutter of media talk that Coe won 3 there before the distribution came out.

    1. Yes there was a big swing away from Labor in Yerrabi which left neither major party with enough of a remainder to threaten the Greens. Some of this went to the Liberals, on account of it being their leader's seat, some went to the Greens and some went to Labour DLP on account of the LDLP drawing first on the ballot. There may have been other factors at play that I am not aware of.

  4. When the full result data tables come out it will be interesting to see what the 3PP and 2PP were. Also I've seen some discussion of what 7 member electorates would have looked like, which could be figured out from the same data. But otherwise I think the analysis is done, and what a fine job you've done indeed

    1. Also who would win the countback for Barr if he left? With so much of her vote coming from off ticket preferences, would Vassarotti retiring (not expected) lead to someone outside the Greens getting up?

    2. In a potential countback for Barr's seat the fate of about half his votes among the three defeated Labor candidates is known. On a three-candidate basis Ingram starts with 21.0%, Northam 16.2%, Anderson 10.9% and the fate of the rest is unknown. On the assumption that Anderson gets eliminated, it becomes 23.6% for Ingram and 18.8% for Northam. It's not impossible that Northam catches up, given that many of the remaining votes are going through female Labor candidates.

      I would expect another Green to replace Vassarotti if she hypothetically didn't complete the term as about half her votes are either her primaries or votes from the other Greens; most of these would go to the other Greens in the recount, pooling with whichever one outlasted the other, and there's no reason to believe any other specific candidate would get enough to catch up.

    3. Full ballot paper preference data is out now

    4. Yep. I know the answers to some of the hypothetical potential recounts but won't be commenting on any of them any further for a while for potential commercial-in-confidence reasons. Sorry folks!

  5. Five 7-member electorates would have delivered the Greens the same 6 seats, on the same votes. Possibly 9-member seats as well.

    1. Interesting exercise - how many seats are needed before there are more Greens elected on the same votes. Going in the other direction (4 seats, 3 seats) would be interesting too.

      Previously the major parties have opposed 7 member seats presumably as it helps Greens, but maybe that isn't the case and it can be checked with data.

  6. Does intra-party candidate choice also mean that "long incumbency matters less"? Canberra Liberal voters don't have to stray outside the Liberal ticket if the party insists on endorsing a Sophie Mirabella or a Wilson Tuckey. Fresh blood in the assembly without necessarily having to throw out the government.