Monday, August 1, 2016

The Senate: Button Press Week

This week if all goes according to plan, the buttons will be pressed on the remaining Senate races in ACT, WA, SA (Tuesday), Victoria (Wednesday), Queensland and finally NSW, and the makeup of the Senate should be known.  I'll be posting some comments about these races and the results as time permits, but I'm rather busy this week, so feel free to add thoughts about any of the races in comments.

I haven't had the time I would have liked to analyse these races in detail, and really you have to sample preferences in scrutineering to have a really confident handle on what's going on in a given state under the new Senate system.  You also can't do it on the night, as you have to know what the main races are to know what you need to look for.  I used scrutineering-based modelling to successfully predict the result in Tasmania (though there was very nearly an upset for the final seat) and it surprises me that I have not seen any detailed public attempt at a scrutineering-based model for any other state.

Some Findings From Tasmania

Tasmania is a little bit different, because it has a massive below-the-line voting rate (29% in this case, including 10% for two major party Senators who were demoted in preselection).  Also Tasmanian voters have a lot of experience with proportional representation, and they don't have much experience with following how-to-vote cards (which are banned on state election days).

Nonetheless there are a few patterns from Tasmania that are worth watching to see if they crop up in the states that are concluded this week.

1. Low how-to-vote card follow rates

Above-the-line voters in Tasmania just barely followed how-to-vote cards at all (see here for rates).  Only the Liberal Party card managed a double-figures follow rate among that party's voters (barely) with only 2.8% of ATL Labor voters following the Labor card and some micro-parties down as low as 0.1%.  Rates in the Northern Territory were somewhat higher, but still below 20% in all cases (hat tip: Dean Ashley).  Rates in other states don't seem to have been exceptionally high either (at least, it seems most voters haven't followed the cards).

See also the results for WA below.

2. Preferences splatter

Further to 1, preferences go all over the place and strong preference flows are a great rarity.  There was no case in Tasmania of any party getting over 40% of another party's #2 votes, and only four cases of over 30% (one of which included the donkey vote).

See also the results for WA below.

3. Popular parties get more preferences

There is a very strong relationship in Tasmania between a party's primary vote and its average preference flow from all other parties.

(The biggest outlier there is the Renewable Energy Party, which has a would-you-like-a-pony sounding party name that probably sucks in a lot of preferences without people realising it is actually a wannabe preference-harvester. Also note the dire performance of the Christian Democrats, whose lead candidate was an interstate ring-in, but far from the only one in Tasmania.)

All else being equal, bigger parties will beat small parties on preferences, but things like a right- or left-wing bias in preference sources can even that up.  Often the right-wing micro-parties have higher vote shares than the left-wing ones.

Because of this feature, we need to keep an eye on any popular party that has even a small fraction of a quota left over, in case its higher preference share allows it to pass all the micros.

4. Ticket position affects preferencing

... but not that much.  There's a slight tendency for voters to preference parties that are close to the party that is the preference source, so that being very close to an excluded candidate might be worth about a 10% better preference flow than otherwise (eg 20% might become 22%) and being far away from an excluded candidate might be worth about 15% worse (eg 20% might become 17%).

Through the count a party will get its chance at being close to some parties and far from others, so this might tend to cancel out.  Nonetheless a party could be moderately advantaged by being just downwind (to the right) of several juicy preference sources.

The Races

As the buttons are pushed and results come out I'll be posting comments for each race on this thread.

Based on the leads on primary vote quotas, if there were no preferences the outcome of the election would be 30 Coalition 27 Labor 9 Green 3 One Nation 3 NXT 2 Liberal Democrat, Jacqui Lambie and Derryn Hinch.  

However Bob Day (Family First) has beaten Labor in South Australia and been returned, so Labor are now only on for 26.  Also One Nation has won two seats at the expense of the LDP in Queensland.  If NSW goes according to the primary leads then the Coalition have these two pathways to pass legislation opposed by Labor:

* The Greens.
* Nine of eleven crossbenchers: all of One Nation, all of NXT and any two of Leyonhjelm, Lambie, Hinch and Day

It's also possible - perhaps even probable - One Nation will fragment (again) or not vote as a block, bringing new pathways into play.

However there is some chance more of the leads could be overturned on preferences.  The most intriguing one now (edit: and yes, it's happened) is Queensland where One Nation's second candidate is way behind (0.1876 quotas - maybe more like .186 after leakage - plays the Liberal Democrats' .3709) but One Nation are likely to get strong preference flows because they are the more popular party and were high-profile in the Queensland campaign.  I am not sure how much detailed scrutineering has been done on this one.  I have seen some sampling which is at least consistent with the idea that One Nation are competitive for a surprise fourth seat, but I don't think more can be said than that at this stage (at least not by me). Any insights welcome in comments (noting that, for various reasons, comments may take a few hours to clear over the next few days).

In Victoria I don't expect anyone to catch the Greens (they have the advantage there that Labor will be out of the count quickly, meaning a lot more preferences for them from the left-wing micros). I initially didn't expect anyone to catch the Coalition, but after Bob Day's performance on preferences in SA it could be that micros, especially One Nation, are a threat to them.  In South Australia it seems possible Bob Day will catch Labor, as he has been reported to be getting 54% of the Liberal preferences (though that's not enough by itself) but he is a long way back and would have to overturn the trend in other states of small parties doing poorly on preferences (edit: he's done just that).  In New South Wales we have one micro party against another (LDP vs CDP) with the LDP starting with a handy lead.  In Tasmania the Christian Democrats performed remarkably badly on preferences and it remains to be seen if that is also so elsewhere.

Western Australia

5 Liberal 4 Labor 2 Green 1 One Nation

The button has already been pressed and an expected close race between the Greens and Nationals failed to materialise with the Greens winning by over 25,000 votes, consistent with the idea that parties that do well on primary votes do well on preferences too.  The exhaust rate at the point where the contest was determined was 3.6%, increasing to 6.3% only following the irrelevant throw of the Nationals candidate to determine order of election for Rod Culleton and Rachel Siewert.  This throw also caused both these candidates to finish over quota, but that is again a meaningless figure; they were both below quota at the time their election was assured.

As widely noted the elected One Nation candidate appears to be ineligible so we can expect a dispute of return at some stage and if it succeeds then based on past precedent, the seat will go to the second One Nation candidate, assuming it is found that Culleton was ineligible at the time of his election.

WA How-To-Vote Card Follow rates: These are the percentages of voters following the how-to-vote card from 1-6 as shown by the ABC, with thanks to David Barry's preference explorer again: Liberal 32.2%, Aus Christians 15.9, Labor 15, Green 10.3, Aus Liberty Alliance 9.9, Rise Up Australia 6.0, Socialist Alliance 3.4, One Nation 2.1, Shooters Fishers and Farmers 1.6 (gave two choices for sixth), Animal Justice Party 1.6, Cyclists 0.7, Family First 0.6, Liberal Democrats 0.4, Hemp/Sex 0.2, Renewable Energy 0.02.  Apparently just one voter out of 4114 REP voters followed the card all the way through (damn that person!) though the figure is so low I wonder if they might have changed their card.

WA Preference Flows: The strongest preference flows seen in WA (#1 to #2 ATL, whether the votes were actually transferred or not) were National to Liberal 54.4%, Liberal to National 50.7, Labor to Green 48.4, Aus Christians to Family First 36.4, Liberal to Liberal Democrat 35.2, Socialist Alliance to Green 34.8, Green to Labor 30.8.  These are the only 30+% 1-2 preference flows.

Quite a few WA micros preferenced One Nation above any other party; these were Australian Liberty Alliance, Hinch Justice Party, Rise Up Australia, Australia First and Mature Australia.

WA Preference Share graph:

Here's the same graph of top-6 preferences as I posted above for Tasmania (data again from Poll Bludger):

Again a strong relationship between vote share and preference-attracting ability from other parties.  The Greens had an extremely high average share (hence their runaway victory), though even a share typical for their vote size would have been easily enough.  Again we have the Christian Democrats performing extremely poorly for their vote share and the Renewable Energy Party very well, but one notable difference is Family First doing very well relative to their vote size in WA.  This seems to have also been the case in South Australia and Victoria, but not in Tasmania, and I'll take as much credit for the latter as anyone out there would like to give me.  (Actually more major causes are probably the low HTV card follow rates in Tasmania and the FF primary being inflated by the donkey vote.)


1 Labor 1 Liberal

As expected of course.

South Australia

4 Liberal 3 NXT 3 Labor 1 Green 1 Family First

After opposing the system claiming it would disenfranchise his voters, Bob Day has been re-elected under it, overtaking the third Labor candidate by genuine voter choice!  I'm very busy today but sometime tonight will unravel how he did it.  (Actually, see the comments by Alaric).

The exhaust rate in South Australia was a mere 2.0%, the lowest in a state so far!

SA How-To-Vote Card Follow Rates: Following card follow rates in SA is slightly tricky because it appears that the major parties each had at least two cards.  There are other orders that look like cards but appear to actually be mistaken attempts to follow cards where a certain mistake is often made, like confusing columns O and Q.  (Queensland will be much, much harder - there was a lot of chopping and changing there.)  Anyway I get the follow rates as: Liberal at least 29% (at least two cards combined), Family First 12.4% Labor 12 (two cards combined), Aus Liberty Alliance 7.6, Green 5.3, Shooters and Fishers 3.9, One Nation 2, Animal Justice 1.8, Lib Dems 1.4% (per card on website, nobody followed card on ABC site), Cyclists 0.4, Sex/HEMP 0.3.

SA Preference Flows: The strongest preference flows seen in SA (#1 to #2, whether distributed or not) were:  Christian Democrats to Family First 57.5, Mature Australia to Labor 43.6 (mostly donkey-votes), Liberal to Family First 43.04, Labor to Green 39.5, Green to Labor 34.8, ALA to One Nation 34.4, Arts to Green 32.8, Family First to Christian Democrats 31.3.  There were no others above 30%.

It seems there is a lot of consistency between states with the rate of ATL voters stopping before six preferences; this has run between 5% and 6% in each state so far.

Antony Green has a thorough summary of the SA Senate cutup.


No surprises here, it's 5 Coalition 4 Labor 2 Green and the one and only Derryn Hinch.  Grahame Bowland has reported that Hinch gets a six-year term in his S282 recount, which is certain to spice up that little debate (since Hinch doesn't do so by the order-of-election method!)

There is a mild surprise in the exclusion order: Family First was the last party excluded, rather than the more fancied Sex Party (second last) and One Nation (third last).  The margin between Hume (Coalition) and Bain (Family First) was 26,812 votes, or .0996 quotas.  This one had one of the higher exhaust rates so far (5.2% at the point the result was determined at, rising to 8.6% after the throwing of preferences that determined only the election order.)

Victoria How-To-Vote Card Follow Rates: Victoria had the highest follow rates so far for the big parties, but that's not saying much.  Coalition 39.7, Labor 17.8, Green 16.2, Rise Up Australia 8.3, Aus Christians 5.6, Socialist Alliance 3.8, Animal Justice 3.4, One Nation 2.9, Sustainable Australia 2.5 (card went to three only, so this is pitiful), Sex/HEMP 2.1 (much higher than in some states), Shooters Fishers and Farmers 2.0, DLP 2.0, Aus Liberty Alliance 1.3, Family First 1.0, Liberal Democrat 0.9 (two cards), Marriage Equality 0.8, Pirate Party 0.3, Science/Cyclists 0.2.  And beating their own WA record were the Renewable Energy Party with 0.01% (1/7981). Double damn that one person! (The reason REP do so poorly is that their HTV was not easy to find on their website.)

Trumping all these is an apparent 51.6% follow from PUP to AJP, Group B, Labor, Science/Cyclists and Derryn Hinch in that order.  These are the first six parties on the ballot.  It seems this was a card but the strength of its flow for a small party is mysterious.

Victoria Preference Flows:  There were only three flows of over 30% from 1 to 2 in Victoria: Palmer United to Animal Justice 54.5% (card assumed), Liberal to Family First 47.8 and Liberal Democrat to Coalition 30.3.  Some parties did not have even a 10% flow to anyone and Group B distinguished itself by having its leading preference flow as exhaust (ie just-vote-1).

The donkey vote across the page starting with Hinch was 0.18%. 


One Nation's second candidate has chased down and passed all the other micros on preferences so it's 5 Coalition 4 Labor 2 One Nation 1 Green.  Both One Nation and the Greens will be slightly over-represented compared to their vote share because of the preferences they pick up from minor right- and left-wing micros that do not have enough votes to win seats anywhere in their own right.  In the end the Liberal Democrats' performance on preferences in Queensland was so poor they were overtaken not only by Malcolm Roberts (PHON) but also by Family First, with Roberts winning pretty easily.  A higher exhaust rate in this count: 4.2% at the point where the result was determined, rising to 7.7% after irrelevant distributions.

Queensland How-To-Vote Card Follow Rates: I hope I'm on top of all the chopping and changing with the Queensland HTVs, but it's probably impossible.  I have LNP 20.0% following 1-6, Australian Liberty Alliance 18.0, Labor 11.9, Rise Up Australia 8.9, Family First 7.6, Katter and Hanson both 5.2, Lazarus 4.9, Green 4.3 (for card shown on ABC website but there were at least three others used so this is an underestimate - see comments), Sustainable Australia 1.5 (only 3 preferences), Australian Christians 1.4, Animal Justice 1.0, Shooters Fishers and Farmers 0.6, Pirate Party 0.4, Sex/HEMP 0.3, Veterans 0.2, Liberal Democrats 0.2 (only 3 preferences!), Democratic Labour Party 0.09, Cyclists 0.07, and of course Renewable Energy Party 0.02.  (1 of 5531).

Queensland Preference Flows: The strongest 1-2 flows in Queensland were ALA-One Nation 48.9%, LNP to Family First 42.5, LDP to LNP 38.5, Christian Democrats to Aus Christians 35.5, the same in reverse 35.3, Cyclists to Arts 33.3, Labor to Green 31.8 and Rise Up Australia to Family First 30.3.  The donkey vote rate (1-6 ATL across page) was 0.16%.

A peculiarity of the count is that Malcolm Roberts received only 77 below the line votes (I assume this is the lowest ever for a Senate winner), compared to  20,927 for Pauline Hanson. 

New South Wales

No overtakes on preferences here - it's 5 Coalition 4 Labor 1 One Nation 1 LDP 1 Green and David Leyonhjelm is returned.  Leyonhjelm beat the Christian Democrats easily by over 38,000 votes (.11 of a quota).  NSW had the highest exhaust rates so far (7.3% at the point where the CDP were excluded, and 9.2% after irrelevant throws). It also has the highest rate of voters stopping before 6 above the line (9.3%).

NSW How-To-Vote Card Follow Rates:  Stand around and gaze in awe, behold this wondrous artefact:
All this useless beauty
The above is the NSW How-To-Vote card for the Renewable Energy Party (who else).  8008 NSW voters voted 1 for the REP.  The card gave a second preference to Animal Justice, and 414 voters did likewise.  The card then went to Sustainable Australia, and 25 of these voters did likewise.  Preference 4 goes to the Sex Party - and was followed by no-one!

This was almost matched by the Veterans Party, whose card was followed to 4 by a single voter, who then discovered that you couldn't follow it to 5, as it had preferenced the Science and Cyclist parties separately, and they were on a joint ticket.

Rates for parties were Liberal/National 30.4%, Labor 13.7, Green 10.7, Aus Liberty 9.1, Shooters Fishers and Farmers 2.8, Animal Justice 2.6, Sustainable Australia 2.5 (card only went to 3), Socialist Alliance 2.4, One Nation 1.4, Rise Up Australia 0.6, Pirates 0.5, Science/Cyclists 0.4, Sex Party 0.4, Family First, Liberal Democrats and DLP all 0.2, Veterans 0.02 to four (see above), and Renewable Energy 0.

NSW Preference Flows: The following were the strongest flows in NSW: Liberal Democrats to Liberal/National 45, Christian Democrats to Liberal/National 44.2, Voteflux to Christian Democrats 38.9 (these were Liberal/National HTV card followers who put the 1 in column E not column F), Liberal/National to Christian Democrats 38.1, Labor to Green 35.9, Drug Law Reform to HEMP 35.8, Green to Labor 32.9.

The Sex and HEMP parties ran on separate tickets in NSW, but together in other states.  Preference flows between them were Sex to HEMP 12.9 and HEMP to Sex 14.7%

The donkey vote (1-6 across the page) was 0.16%.


David Leyonhjelm's re-election means that four of the previous non-Green crossbenchers have been returned and four have lost.  They've been joined by two new NXT, four One Nation and Hinch so the non-Green crossbench increases from eight to eleven.  My simulations of the 2013 result as a DD had 12 non-Green crossbenchers, so the paranoia that a DD was going to wipe the crossbench (bar NXT) out was always silly.  I didn't actually expect there to be so many crossbenchers this time around, but the way in which other micros (primarily One Nation) filled the void left by the implosion of PUP made that possible.

So the Senate is done and dusted (barring any challenges) at Coalition 30 Labor 26 Greens 9 One Nation 4, NXT 3, LDP 1, FF 1, Lambie and Hinch.  

The next fun will be the Section 282 proposed allocation for 6 and 3 year terms, and whether or not the Senate agrees to it.


  1. Thanks Kevin. What do you think the chances are of a micro-party like the Sex Party (second preference on the ALP how to vote card and helpfully placed right next to the Greens at the end of the ballot) snagging the 12th spot from Janet Rice or the 5th Coalition candidate?

    1. I'd say low but without any scrutineering data in Victoria can't say for certain. If what we saw in Tasmania is any guide the left-micros shouldn't be a threat to the Greens since the Greens will get more left-micro preferences than any of them will. In terms of threats to the Coalition, that also shouldn't be too much of an issue because Hinch will be elected pretty quickly and after that the right-wing micros will ultimately only have the Coalition and each other to go to. Right-micros like One Nation might be an issue for the Coalition though.

  2. Oh, and according to Janet Rice's Twitter feed, the Victorian button press is due at 2:30 on Wednesday. I read elsewhere that SA (along with WA) was due today, but I'm guessing that will be tomorrow. Queensland is tomorrow or Wednesday and NSW will be last on Wednesday or Thursday.

  3. "the Renewable Energy Party, which has a would-you-like-a-pony sounding party name that probably sucks in a lot of preferences without people realising it is actually a wannabe preference-harvester." Too true, Kevin. Breenie's good at that - he got elected to the NSWLC by naming his party "Reform the Legal System". He got it registered by signing up a lot of the people ripped off by the Home Fund, and then got votes on the sound of the name, and a flukey preference deal. I'll pass your comment on to him - I think he'll be amused rather than offended.

  4. I got the same answers for HTV compliance in WA (doing it the hard way because I'm an idiot and didn't think to check if David Barry had worked his magic - at least I learnt some more SQL) with the exception of ASP - I think they also had two options for 5th, taking them up to 2.2%.

    I think the REP result may be real, although they may not have distributed their card very widely! The four most common combinations for them were 1 only (71 votes), 1-to-6 across the card starting from them (29 votes), REP 1 GRN 2 and no other preferences (10 votes) and 1-to-6 backwards starting from them (9 votes).

    I think the interesting outcome is that a few of the micros got some traction. However it seems to be those that actually put out sensible cards versus those who tried to play the game. Compare AUC (preferencing FFP, RUA, CDP, DLP, ALA) to FFP (who put the Shooters second, likely in return for a second preference in SA).

  5. Looking at the results more ALA preferences went to Greens that PHON.

    1. Not what I see (if this means WA). A possible confusion source here is that there are two candidates called Robinson (the surname of the ALA lead candidate) and of those the CEC candidate Robinson's preferences did go more to Green than PHON.

    2. Quite probably I made a mistake.

      Though if I'm reading it correctly now it looks like still 1 to 6 flowed ALA to Green and to PHON.

  6. KB I take it your HTV WA flows is for 1-4 only as per David Barry Table. Is there a Table with 1-6?

    1. My flows are for 1-6. David's page has a downloadable version that runs from 1-6.

  7. Poor result for the Coalition in WA, which is traditionally a conservative state; contrary to what most people think One Nation are only "right-wing" on SOME issues. Based on the 2013 result would they have wanted to restrict the centre-left (ALP+Greens) to four seats?

    1. Yes the 2013 primaries under the new system would have produced something like 6 Coalition 3 Labor 1 Green 1 PUP 1 LDP, so effectively dropping two seats to the left here.

    2. Is that the original election or the separate senate election when the Greens got 15% primary?

  8. The original. The result of the rerun under the new system would probably have been 5 Liberal (or 4 Lib 1 Nat) 3 Labor 2 Green 2 PUP.

  9. Some brief analysis of the SA result:

    - Bob Day started 0.096 quotas behind the ALP #4 after surpluses were distributed, but 0.112 quotas behind combined ALP #4-#6;

    - He cut deficit to 0.069 quotas before the LIB #5 was excluded and had a 0.018 quota lead after that exclusion.

    - His lead expanded to 0.096 quotas on exclusion of One Nation before slipping back to the final 0.043 quotas on the surpluses of XEN #3 and GRN #1.

  10. Confirming that the button press for Qld is 9am on Thursday, with the formal declaration happening Friday at 11am.

  11. A further snippet for the "Bob Day – how did he do it” files.

    In counts 11 to 435, a total of 0.98 quotas was distributed from minor parties and down-ticket candidates to the final six. That split went as follows:

    LIB 11.4%
    ON 15.2%
    ALP 14.0%
    FFP 16.8%
    GRN 18.8%
    NXT 15.5%
    Exhausted 7.8%
    Losses 0.4%.

    These counts do not include the big boost to FFP from the next two counts which excluded the LIB and ON candidates.

    The fact that Day (and One Nation) defied the “big parties get more preferences” trend from Tas and WA to more than keep up with the majors through these counts is what kept FFP close enough to the ALP that the LIB and ON preferences could get Day to the line.

    The other thing that killed the ALP is the expected one – the poor performance of the Greens and their presence in hoovering up left-wing preferences right to the end. For example Day actually gained on the ALP in the counts that eliminated the AJP.

    1. I'm not sure I follow the argument in your final paragraph. If the Greens got a higher vote, wouldn't that have been taken mostly from Labor or the "left micros" anyway?

    2. Bob Day (Family First) would have been lower than the 5th Liberal in South Australia if Robson rotation applied and an ATL vote for a party goes down the team in order on each ballot paper. If that applied, no one gets a quota on first preferences so the major parties do not bleed 2nd preferences from BTL voters who vote across parties. Instead of 10 counts distributing surpluses straight down the party machine tickets, the count would start with elimination of lowest candidates with none exhausting due to 2nd to 6th prefs already elected. Each candidate would go out in one count, because there would be no fractional value papers, due to lack of surpluses above quota.

      The Below-the-Line voters would determine the relative standing of candidates in each team. As a rough assessment, at the time Edwards, the 5th Liberal was eliminated, the 4 above him were elected on full quotas more than 400 counts earlier. Take 11,000 votes off these four and add 44,000 to Edwards. One Nation Party is now lowest, not Edwards. The 5 Liberals probably get enough from One Nation to stay ahead of Bob Day to the end.

      Geoff Powell

  12. I’ve just finished calculating exhaust rates by candidate for Tasmania. Below are the rates for ATL votes, which exhausted at a slightly lower rate (2.5%) than BTL votes (3.6%).

    Parties that Elected a Candidate or were in the final count

    Australian Labor Party 0.1%
    The Greens 0.0%
    Liberal 1.9%
    Pauline Hanson's One Nation 0.0%
    Jacqui Lambie Network 0.7%

    Other Parties

    Family First 12.1%
    Christian Democratic Party (Fred Nile Group) 7.6%
    Nick Xenophon Team 6.1%
    Palmer United Party 12.4%
    Australian Sex Party 21.6%
    Derryn Hinch's Justice Party 11.8%
    Citizens Electoral Council 39.1%
    Renewable Energy Party 16.4%
    Australian Liberty Alliance 20.1%
    VOTEFLUX.ORG | Upgrade Democracy! 13.1%
    Shooters, Fishers and Farmers 16.8%
    Animal Justice Party 13.7%
    Science Party 14.6%
    Australian Recreational Fishers Party 13.3%
    Liberal Democrats 15.8%
    The Arts Party 19.2%

    The outcomes shouldn’t be a surprise. If you don’t get a candidate elected or into the final dance, some votes are going to exhaust. But it still surprised me how much the overall rate depended on the extremely low rates for the contenders,

    Of those contenders, JLN and the Greens are zero because they were there at the end (keeping in mind this is lust ATL votes, so no leakage) and the Liberals are highest because Bushby had a significant surplus after Colbeck went out.

    Everyone else apart from the CEC (only 130 votes) was between 6% and 22%.

    I’ve also posted these in the comments of Antony Greens blog on exhaust rates, where they will probably cause a stir among the end-of-the-worlders. But the lesson is easy – if your horse isn’t in the race, have a place bet on one that is!

  13. "Both One Nation and the Greens will be slightly over-represented compared to their vote share because of the preferences they pick up from minor right- and left-wing micros that do not have enough votes to win seats anywhere in their own right."

    Well, that's the point of preferential voting isn't it?

  14. I may not be so excited about the QLD result from a political spectrum perspective, but from a mathematical perspective I think it shows a very important point. Voter preferences were very important in allowing PHON to go from about 1.2Q on primaries all the way to 2 seats.

    And... it's legit! Form a democratic perspective we can truly say: this is what people voted for. Everyone giving PHON a #2, #3, #4 (etc.) did so with their own pencil and of their own accord, in full knowledge that they did it.

    It's a beautiful thing. From a democracy perspective.

    I don't like the result but I LOVE how we got to it.

  15. Has the AEC provided any timing on the recount on the basis of 6 senators (for information purposes to the new Senate)..
    Presumably this is just another pretty simple button press.

    1. I've seen no information on when this will be done, though I'd assume before the end of the month and after all the seats are formally declared. There are some complex debates about exactly what assumptions should be used but I'd expect the AEC has already taken a view on those and tested their ability to implement them. I hope they will be documented at length. We should at least get uniform standards between the states in how S282 is applied this time which apparently was not the case when it was done manually in 1987!

  16. Well, my longshot 5th ALP candidate in NSW prediction didn't come off. Looks like the ticket leaked a fair bit (down to .045 once all other Labor candidates excluded). I'm still wondering how they would have gone if they could have got ahead of the remaining left micros (HMP, ASXP and AJP). ASXP and AJP had seemingly high exhaust rates as there weren't many appealing options left.

    One interesting thing I noted was the high preference flow from Health Australia Party to Family First. This has to be a donkey/ballot proximity vote, right? I wonder how many of these made their way down to the LDP in ballot position #4?

  17. Has anyone run the numbers to calculate what the result would have been if it was not a DD? I guess the 6 year term calculation should show that approximately.

    1. A complication here is that a half-Senate election on the votes cast would elect Lambie. But had there been a half-Senate election, Lambie would not have been a candidate and it's highly unlikely her ticket would have got anyone up without her at the helm. (Some have argued the same thing re NXT in SA but I'll give them the benefit of the doubt on that one).

      For a half-Senate election it seems to come out as 17 Coalition 14 Labor 4 Green 2 NXT plus Hanson, Hinch and Lambie. Redirecting the Lambie votes gives Labor a 15th which goes to Lisa Singh.

  18. Just in regards to your assessment of the numbers following the Greens how to vote cards in Queensland, it is worth noting that in that state the Greens had about 4 variations in the Senate component of their how to vote cards in different electorates - fairly minor in the scheme of things and nothing that would have influenced the final result but certainly sufficient to mess with statistical analysis of how closely the how to vote was followed.

  19. I think many voters may be influenced by how to vote cards without actually following them. Rather, they'll look at the sheet their given and use it as a guide, making adjustments. In particular they might have a strong idea about their number 1 and 2, but are more likely to look at the HTV for the last couple of spots to round out their 6. Others might look at the HTV for their entire 6, but re-order it based on their own ideas about those parties, without actually looking at others. Particularly I'd bet that a lot of NSW voters couldn't have been bothered to read all the party names.

    So, I don't think checking for the complete sequence in order is necessarily the best way to gauge the full influence of a HTV card. Would it be possible to instead determine the average preference number for each party on another's HTV card amongst their voters?

    1. I've been using David Barry's preference explorer (loading up all the votes in most states without getting issues is beyond my computing abilities at the moment) so I don't have the ability to determine the average preference number.

      What you suggest would have to be true to some extent, but I think that determining that extent from the data is extremely difficult. Even if there is a fairly strong match between the HTV and the average preference numbers for a party, it could just mean that the party has preferenced parties that its supporters would be inclined to preference anyway.

      There are some things I do notice though, that suggest that the influence of the cards is genuinely very small at any level whatsoever for most parties. Firstly, if a party preferences a likeminded but obscure party second, then the proportion of voters that preference the obscure party 2 tends to be very low. So that suggests those cards are not influencing voters to do anything they wouldn't have done anyway. There are plenty of cases where the party put 2 on the card is way down the list of parties most commonly placed 2 for the preferencing party. Secondly there is a quite strong aversion among the rural-right parties to preferencing the religious-right parties even when the card recommends this.

      What we're seeing with all the small parties here is partly that they are unable to influence preference flow because they lack the means to do so. It seems voters don't look up how to vote on the party websites to any large degree so if no-one hands out at the booth then the voter does what they like. It also seems that if a left-wing micro party hands out a card, nearly all its voters will totally ignore the card anyway, and direct preferences to Labor and the Greens rather than whatever left-wing micros the card says.

      When it comes to the Coalition (who have by far the highest 1-6 follow rates) the average strict follow rate (1-6) for the five largest states was 30.3%. The average follow rate to 2 was 44.4%. Of the 14% following to 2 but not to 6, I'd suspect quite a lot weren't influenced by the card at all, so I suspect the percentage being loosely influenced isn't massive. I also notice that the proportion juggling the HTV card order to the extent of putting a party from 3-6 on the card second is often small, eg in Queensland the parties 3-6 on the LNP card had just under 10% of the LNP #2s between them, which they would have likely got anyway.

      Something else I noticed quite a bit is that a voter who follows a big-party card to 4 will nearly always follow it to 6. For instance in Queensland 94.5% who followed the Coalition card strictly to 4 followed it to 5 and 6 as well. Only 0.2% of those who followed it to 4 then had the last two parties on the card in reverse order.

      Big parties might have done some experiments on this. It's possible for instance to hand out a different card at a specific booth then see if that booth behaves differently.

    2. Anecdotally, I can say that I looked at a few online HTVs for suggestions. But then I filled in ATL to preference 19. The HTVs probably changed the order of my 2-10 preferences. If there were other voters similar to me, I'm not sure how this would show up in the data. If a voter for party A is more likely to preference party B than the average voter, is that due to the HTV or just due to innate voter preferences of party A voters. In general, how do you tell if the HTV influenced the voter, or if likely voter behaviour influenced the HTV? Or are they both just caused by political preferences? Maybe next election, the parties issue HTVs based on the preferences their voters expressed last election.

  20. Very belated comment I know, Kevin, but re the REP preferences I didn't find their HTV hard to find on the website - I just thought it was ridiculous, still showing signs of Breen's palliness with Druery and preferencing some very odd parties. I can readily imagine that others who looked it up felt the same. And of course there would have been many who didn't look it up and Made. Up. Their. Own. Minds. As voters are supposed to do.


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.