Sunday, April 14, 2019

NSW Legislative Council 2019: Button Press Day


Button press from 11 am Monday

Won off raw quotas: Coalition 7 ALP 6 Green 2 PHON 1 SFF 1
Coalition will win 8th seat

Labor 7, One Nation 2, CDP, LDP, AJP, KSO for final 3 seats

Labor very likely to win seat, One Nation likely but not so clear, CDP/LDP/AJP who knows, KSO remote chance only

Result

The result will be added here once known.  The count is expected to take a gruelling 45 minutes to one hour.  See https://twitter.com/nsw_upperhouse for updates and doubtless other #nswvotes sources.



Update 11:40: This is going rather slowly so far, with Mark Latham's surplus (4th elected) only now being distributed.  It should speed up when all the ticket leaders are elected.

Ben Raue live stream at https://www.pscp.tv/w/1PlKQLWrpRvxE

12:10 Lead candidates of some of the minor parties are now being excluded.

12:15 Keep Sydney Open are out.

12:18 Labor, Animal Justice and One Nation have won!  Christian Democrats and Liberal Democrats have been defeated.  Animal Justice's Emma Hurst overtook One Nation as well as CDP and LDP.

Analysis

Firstly it doesn't seem we've had any close critical exclusions.  Keep Sydney Open dropped out 8593 behind Liberal Democrats after the exclusion of Sustainable Australia.  By this stage Animal Justice had already passed Liberal Democrats and were just behind Christian Democrats.

The Keep Sydney Open preferences flowed a lot like the 2015 Greens preferences - 20.9% ALP. 12.2% AJP, 4.8% Coalition, 2.6% LDP, 2.3% PHON and only 0.8% CDP.  Leyonhjelm was here excluded (my tweet saying "Leyonhjelm excluded" is my second most popular of all time) 5780 behind Christian Dems, which means he actually went backwards compared to them in the preference flow.

Now the Christian Dems needed to get 15.5% points higher off LDP than AJP did, or 4.0 points higher than One Nation, but they didn't even get 4.0 points at all; in fact they only got 3.3 points thereby sealing their fate.  The margin was 5408 votes to One Nation.

Mark Latham's below-the-line votes were leaky as expected.  50.5% leaked to other parties immediately, and 6.7% went to candidates other than One Nation's #2 Rod Roberts.

More comments soon.


Pre-Button Press Comments
Time for a new thread to bring the contest for the final three seats in the NSW Legislative Council 2019 to the top.  Over the past few weeks I've tried to follow this count on what has become one of the longest threads on this site.  It's the first time I've had the time to attempt to model a NSW upper house postcount in detail - thanks largely to the lack of Lower House seats going down to the wire.  Unfortunately I've made the odd rookie error along the way in doing so and at a certain stage I developed a nagging feeling that the chess maxim "long analysis, wrong analysis"*  (not usually a problem in my case!) applied to aspects of nearly 10,500 words of effort.  As a result, though I never called any of their outcomes as done deals, my outlook was for much of the count too bullish for the Liberal Democrats (who may still win) and too downbeat about the prospects of the Christian Democrats and Animal Justice Party (who might yet both lose, though I doubt it.)

The count has been difficult to model because the NSWEC conducted two separate counts, the "initial count" (a rough throw including above the line votes for seven specified parties, blank informals, with all votes supposedly none of these classified as "Others") and the "check count" in which votes were processed to candidate and party level as they were entered into the NSWEC's computer system.  For most of its history the check count was obviously biased in favour of inner-city electorates and against rural and regional electorates, and it also started with a massive bias caused by the iVotes all being uploaded at once.  (This voting method, being favoured by younger voters, would be likely to create skews that went beyond geography.)

In trying to estimate where parties currently were through the count I assumed the far more complete initial count would be accurate concerning how many votes were left for Others parties, after adjusting for likely levels of non-blank informals and evidence concerning the rates of below the line voting for the seven selected parties.  Initially this worked well at correcting the rampant over-reporting of Others parties in the check count.  However at a certain point, with the count still carrying a clear urban bias, it started saying Others parties were under-reported in the check count.  This seemed plausible if, for instance, absent votes that tended to favour left-wing micro-parties were also under-included in the check count, especially as only some micro-parties have votes very concentrated in the inner cities.  Even so it didn't feel right.

It turned out that the explanation was errors in the initial count.  Every party included in the initial count was under-counted in it with a percentage of its above-the-line votes mistakenly counted as Others.  Not by much (only 0.76% overall) but by enough to bump down all the Others parties significantly.  But more, the two initial-count parties fighting (and apparently struggling) for a place from less than a quota were much more undercounted (over 4% for CDP and over 6% for AJP) than anybody else.  This makes sense in retrospect - the counters would rarely see votes for those parties and would be more likely to forget that they were one of the specified parties and hence mistakenly toss their votes into the Others pile.

And also, the non-blank informal rate increased sharply over 2015.  It was projecting to drop based on areas with high 2015 informal rates being overcounted, but didn't.  Throw in a degree of natural tanking by the Liberal Democrats as the rural booths went in (but probably more tanking than would have been expected) and you have the perfect storm that dragged them from a very good looking position (but still one that should not have been claimed won) to a very nervous one.

A very big thanks to Ross Leedham for his excellent work on Twitter on this count, especially in projecting the Keep Sydney Open vote.

Where They Stand

The following are the primary vote totals for all parties in percentage terms with quotas.  They are ordered by remainder after the number of full quotas.

Coalition 32.82% (7.6597 quotas)
Labor 29.69 (6.5313)
One Nation 6.90 (1.5170)
Christian Democrats 2.28 (.5008)
Liberal Democrats 2.18 (.4794)
Animal Justice 1.95 (.4286)
Keep Sydney Open 1.83 (.4029)
Sustainable Australia 1.46 (.3218)
Voluntary Euthanasia 1.06 (.2322)
Shooters Fishers and Farmers 5.54 (.2182)
Small Business Party 0.68 (.1503)
Greens 9.73 (2.1401)
Australian Conservatives 0.59 (.1300) (just give up Cory.  I mean, seriously ...)
Flux 0.36 (.0801)
Socialist Alliance .0.32 (.0702)
Group L (Buckingham) 0.26 (.0583)
Group G (Seniors) 0.15 (.0323)
Advance 0.09 (.0194)
Group J (Jansson) 0.07 (.0159)
Eight ungrouped candidates combined 0.045 (.0100)
Group H (Monaghans) 0.01 (.0016)

Preference Distribution

When the button is pressed it will conduct a computerised distribution of preferences mimicking what would have happened by hand in the old days, but much faster.  Ludicrously, the process includes random selections of which ballot papers are passed on as preferences - an unnecessary provision in the computer age and one that could produce the wrong winner in the case of a margin under a few hundred votes - but something that is somehow embedded in the constitution.  This wrinkle aside, it's similar to what now happens in the Senate, except that the exhaust rate is much higher.  At this election the proportion of voters expressing at least one preference above the line beyond number 1 increased from 15% to 28%, and this rate tends to be higher for small parties, but something like half the preferences thrown are still going to exhaust (if not more).

As a result of exhaust, you don't need a full quota to win the last four seats this year.  What you need to do is stay out of last place at each exclusion until there are five parties left, at which point the party in fifth place loses and the four above it win.

The relevant general patterns for optional preferencing are:

1. How to vote cards affect flows from major parties when their last candidates are excluded, but have little impact otherwise.

2. Preferences go all over the place.

3. All else being equal, preferences flow more strongly to a party that is politically similar to the party that has been excluded.

4. But also all else being equal, and often overriding 3, preferences flow more strongly to high-profile parties than low-profile ones.

5. Voters for left-wing minor parties including Greens are more likely to use preferencing.

What might occur

The Coalition are in the lead and need to only avoid being passed by four other parties.  Labor might pass them though I doubt it; I greatly doubt anyone else will.  Some time ago I made the eighth seat (Wes Fang) an assumed win; in a great loss to political entertainment Peter Phelps (#9) is out.

Labor are in second place and will win unless three parties below them overtake them.  Their lead is very small and they will lose about 0.01 quotas on leakage from below the line votes, but should still be in front of the others after that.  Based on the past history of preferences flowing to the majors in these counts, and based on the left-wing skew of the available preference sources, I'm not quite calling it in advance of the outcome, but I'll be very surprised indeed if Mark Buttigieg doesn't get up.

From this point on things get quite messy.

One Nation's second candidate Rod Roberts is in theory in third place, and will win provided that not more than one other party overtakes him.  But that theory only holds to the extent that votes from Mark Latham flow through to him.  Of the 0.5133 quotas that will be Latham's surplus, about .03 quotas will be Latham's below-the-line votes.  Latham is a high-profile candidate and quite a lot of voters could have voted for him personally and then sprayed their preferences (also the case with Pauline Hanson and Jacqui Lambie).  I'd guess about half of Latham's BTLs might leave the One Nation ticket, putting Roberts back to about the level of the Christian Democrats.

From there one friendly source for Roberts is the Shooters, Fishers and Farmers.  At the Senate election One Nation was the commonest destination for above the line SFF #2s (nearly 15%, compared to 4-5% for Christian and Liberal Democrats).  But many of the Shooters preferences will exhaust, so this might only be enough to recover the damage on Latham's leakage.  If Senate patterns are repeated, One Nation should also gain a few percent of a quota over the LDP and CDP from the Voluntary Euthanasia Party and Sustainable Australia Party.  That seems like enough if it happens, but I wouldn't be totally sure.  There's a theory that a lot of voters may have voted through above the line specifically to put One Nation last (as a few dedicated souls did below the line).  I'm doubtful of how much impact that will have, because most of these votes will lodge with Labor or the Coalition rather than making it to one of One Nation's opponents at the end.  Still while Roberts' position looks fairly good, I have more doubt about him than Labor.

Assuming Labor and One Nation both get up, the last seat is between the Christian Democrats' incumbent Paul Green, the Liberal Democrats' former Senator David Leyonhjelm, the Animal Justice Party's Emma Hurst and, as what I think is a very remote shot, Keep Sydney Open's Tyson Koh.  I don't think any of the others have realistic chances.  Koh needs to pass Hurst by the end of the expected exclusion of the Animal Justice Party, which means he needs to gain at about 1.7% of a vote for every vote thrown.  That will be a few percent of a vote for every preference that doesn't exhaust.  There's no reason to believe KSO will gain on AJP off the Sustainable Australia votes (different geographic distribution, different demographic too) so they would probably have to do it mostly off the Greens, which is also doubtful.  Even if KSO get over AJP it may be hard for them to pass both the right-wing micros ahead of them from so far back.

In the last count of this type, the Christian Democrats were the slowest-moving competitive party on preferences, gaining only .015 quotas, while the AJP gained .065.  In the 2016 Senate count, the Liberal Democrats gained preferences 45% faster than the CDP.  The preference flows this time could be about 2.5 times stronger as in 2016 as there are nearly twice as many ATL preferences and the total preference pool up to the first key point is 8.6% compared to 6.7%.  However, the AJP's best source last time (Greens) has not quite as much of a surplus this time as last.  If the CDP drop out first their preferences will give the LDP a little bit of a hand,

A lot could come down to the unknown preferences of Keep Sydney Open, which may help the LDP against CDP by being close to them on the ballot paper and by having some ideological common ground, though I wouldn't think that much as I think KSO's voters are typically going to be young and left-leaning.

I would have liked to have time to do a detailed model of this count to form a clearer view on which of these parties might win.  It could be awfully close between LDP and CDP as to which drops first, and I think that AJP are more than just an outside chance - but none of them should be confident! We'll see ...

(* The meaning of this expression in chess is that when an analyst analyses a position by listing a single long series of moves that could have happened from it, there is a high chance that the variation does not represent the best possible play.  The longer the sequence the more the chance that some move in the sequence given by the analyst is not the best move, rendering the rest of the sequence incorrect.  In election modelling, if a model depends on many assumptions then an error in even one of them may ruin the model, depending on how sensitive the rest of it is)

4 comments:

  1. Thanks for your excellent coverage!

    Definitely had a bit of luck on the KSO projection. I even had some doubts when KSO had ~52K votes and AJP had ~44K in the check count. KSO then promptly went nowhere from that point on.

    ReplyDelete
  2. I am crying for those who have no one but us to speak up for them. Congratulations to Emma Hurst. The beginning of a huge change across the planet. Animal Justice parties springing up all over the planet. Goal us ScoMo. One will follow after the other. We will not forsake them.

    ReplyDelete
  3. Not that it's necessarily the most useful analysis, but what would you estimate this result translating to at a Senate half election?

    ReplyDelete
    Replies
    1. Two Coalition, two Labor and one Green, with a close fight between the third Coalition and One Nation for the final seat, which the third Coalition would probably win.

      Delete