Sunday, February 1, 2015

Queensland: post-election wrap and postcount comments

Expected seat outcome: 44 ALP 42 LNP 2 KAP 1 IND

This is my thread for Queensland post-counting comments.  For the first week it probably won't be updated more than daily, if that, and probably won't be detailed.  After that if there is any life left I'll have a more serious look at remaining seats.

Thursday 12th 5:00: One of the few remaining close seats, Lockyer, has been declared, with Pauline Hanson missing out again, this time by a feeble 114 votes.  Non-exhausting Labor preferences split over 58% in Hanson's favour against the LNP.

Thursday 12th 3:30: What little chance existed of the Ferny Grove result being wiped has pretty much gone out the window with news that the number of votes exhausting at the exclusion of the allegedly ineligible PUP candidate is lower than Labor's margin in the seat.  This means that the argument that "had all those who voted 1 PUP and stopped instead voted 1 LNP the result would have been different" does not in fact apply and any remaining substance to an appeal will likely be disposed of based on precedent.

Thursday 12th: The LNP have survived by 126 votes in Mt Ommaney; had Labor won this seat it would have a majority for Labor.  Seats are being progressively declared, with most done now, and soon the election will be over.

Tuesday: I've added nothing more because there's nothing to add; all the leaders are keeping their lead and the expected outcome is done and dusted unless someone finds 500 votes under the sofa.  As noted by Antony Green, the LNP cannot try the usual gimmick of the defeated Premier going back to the House to test his support, because the defeated Premier is not a member of parliament, so it should be that Annastacia Palaszczuk is about to be commissioned as Premier of Queensland.  See also now Pollbludger.

Sunday: Chris Foley is too far behind in Maryborough and will be eliminated based on an indicative throw; Labor will win the seat.

Saturday: There are confused reports about whether Chris Foley has conceded Maryborough (he's apparently conceded he's unlikely to win, which is not the same as conceding he's definitely lost, not that concession makes any difference anyhow).  It appears that during an indicative preference throw he is about 900 behind Labor with about 4000 preferences (presumably from PUP) to add, which is a hopeless position.



Friday: I'm looking for reasons for realistic doubt that Labor has won this election, and I can't find any anymore, bar the outside chance for Foley in Maryborough.  Re Ferny Grove and a possible challenge, see Antony Green's excellent coverage.

Thursday: Margins in various seats keep bobbing up and down with Mt Ommaney and Lockyer now the closest.  However none are micro-close at this stage and it may well be there will be no changes and that Labor is set to govern with support from Peter Wellington, whose support was announced today.  Note that if Ferny Grove is disputed, the dispute will take months to resolve as it will be a significantly novel Disputed Returns case.

Wednesday: Really not a lot has changed today so I have just made some minor changes to the list at the top.

Tuesday evening:  Gaven a non-issue now. There is a possibility of a challenge in Ferny Grove where a PUP candidate was apparently ineligible to contest the election by virtue of bankruptcy.  It might be argued that had those voters who voted 1 PUP and then exhausted their ballot not had that candidate to vote for, they would have instead voted 1 for another party, which might have altered the outcome.  It's a long bow to draw though, even if it turns out to be mathematically possible (which is unclear on present figures).  In any case it's likely the usual practice would be followed: the seat result would be declared, and there would then be a wait to see if there was a challenge and hear it if so.  Only after a successful challenge would there be a by-election.

Tuesday morning: Ross Irwin has a long post in comments about Bundaberg that suggests the margin has been coming down rapidly in this seat, causing me to include it in the list of seats at some level of doubt, along with Lockyer. Also while I expect the LNP to beat Hanson I have transferred the seat to the unresolved list as I think the level of doubt about the projections involved there is a bit above token.  The celebrity political ghost rises to probably just fail in yet another postcount!

ECQ practices of removing indicative preference throws while reallocating votes are making the count difficult to follow even for those not posting from Corinna.

Monday updates: Maryborough remains unclear with Chris Foley now needing to gain 1192 off 6247 minor party preferences over Labor (a gain rate of .19 votes/vote) to beat Labor into second and get his old seat back.  This would be much more achievable with compulsory preferences but with the votes being relied on coming more than half from PUP with most of the rest from One Nation, it sounds like tough sledding because of the exhaust rate.

There was a flutter of interest in Lockyer where Pauline Hanson is on 27.3% and will make the final two if Labor cannot gain 2.38% on 14.11% of KAP, PUP and Green preferences (KAP being the big share).  (The gain rate here is .169 votes/vote and Labor should gain strongly off the Green votes, so it isn't quite clear Hanson is second after preferences yet - that said the KAP votes could be handy for Hanson.)  Current 2CP counting has Hanson leading the LNP if she is second, but this is based off a limited and unrepresentative subset of booths.  William Bowe reckons there's not going to be anything to see here by the end - I've not had time to crunch it in detail yet.

Gaven has also sprung to some sort of life for reasons noted by William (the 2CP in this seat originally picked the wrong candidate as the incumbent independent lost heavily - now based on preferences the LNP are projected as winning only narrowly.)

Developments in the other classic-2PP LNP-ALP seats can be followed on Poll Bludger.  As has been noted 44 seats gives the ALP government with support from Peter Wellington.  Whether 43+Foley+Wellington would also do the trick I do not know; KAP have indicated they are ready to consider supporting either party but that the LNP would have to dump proposed asset sales.

=======================================================================

Wrap: What The Hell Was That?

We'll be waiting a little while to be certain but Labor is probably back in Queensland only a term after being thrown from office in the most disastrous result in seat terms in Australian electoral history. It may or may not be in majority but in practice it will barely matter which.  As I write the LNP has only a slim chance of holding enough seats to have something to seriously play for at the negotiating table.  Premier Campell Newman has been unseated and it looks likely the electorate has spared his party the mess of choosing a new Premier by choosing a Labor Premier instead.

In terms of the 2PP picture (currently looking like approaching 52:48 as expected, but to the other side!) this is unfortunately the bit where I get to fulfill the definition of a psephologist as one who spends the day after the election explaining why their expectations were wrong.  Excluding those polls that have obvious house effects (of which there aren't many in Australia at the moment) there are three kinds of errors polling-based projections make.  The first is if the primary votes are different to what the pollsters expect, the second is if the assumed flow of preferences is wrong and the third is if the method of converting the expected 2PP to seats is incorrect.

In both the federal and Victorian state elections both the first and second errors happened, but they cancelled each other out: changes in preference behaviour favoured Labor but were modest, and the primary votes were better for the Coalition than expected, by a similar amount.  In this case both the errors have gone in the same direction: both Labor and the Greens polled a point better than expected (at the expense of Others) and there has been a massive change in preferencing behaviour.  The first issue is worth about a point in Labor's favour on current primaries while the second is worth about 2.5-2.8 points (we'll know exactly how much when we see the final figures.)

According to Poll Bludger the change in preferencing saw the total spread of minor party preferences change from 27-22 to Labor (51% exhaust) to 45-15 (39% exhaust).  On a two-party basis Labor's gain rate per minor party vote changed from .05 votes to .30 votes.  While a change in preference behaviour was to be expected, this magnitude generally wasn't.  After years of scoffing at respondent-allocated preferences because of their history of failure (especially 2004) we've finally had a case where last-election preferences were not just modestly wrong but quite off the planet.  Respondent-allocated preferences probably would have done much better.

But that's not easily testable, because as far as I know, none of the public pollsters released data on respondent-allocated preferences for this election. Now this is possibly because those releasing results calculated exclusively on respondent preferences get told off for doing so by people like me (allegedly excessively), especially if the choice of that method surfaces in a commissioned poll that happens to have results that suit the agenda of the group releasing it.

But following this result we will have to look much more closely at whether massive changes in preferencing behaviour like this one in Queensland can be more reliably predicted - especially under OPV where there is the extra issue of who will choose to distribute their preferences.  An obvious starting point is that share of minor party preferences is probably correlated to major party primary votes and that past preferencing behaviour should be less reliable and possibly not used at all when the swing is very large.  NSW will be a further test of this. We need to encourage more pollsters to release both last-election and respondent preferences (as Morgan and Ipsos do) and then those using the data can take our picks of which one we trust the most.  I'm suspecting that at federal level - where a shift as dramatic as this one is deeply unlikely given Labor's already higher base - it may work best to use a hybrid measure of the two.

I'll have a review of pollster accuracy when the final primaries and 2PP are known.  Obviously, no-one was all that close to the pin.

Governments That Lose The Plot

I have a general feeling following that the Newman government, the Abbott federal government and the Rudd/Gillard Labor governments have all got it wrong in roughly the same way.  The Hodgman Tasmanian government has to be careful to avoid the same mistake, which it has been showing some signs of making.  Governments that win from opposition have recently taken it as their mission to take drastic economic steps and to engage in culture wars to erase their predecessors' legacies and appease their base (or for Gillard, crossbench supporters).  Voters in my view don't want this: they want solid, stable boring governments that aren't obsessed with erasing debts immediately or appeasing groups out of the mainstream.  They want governments that don't promise them the stars and then give them something entirely different.

As for Abbott, it is my belief that this Queensland result, plus the current flood of leadership speculation and very bad polling (with worse likely to come), is terminal.  The Coalition should not be complacent about retaining office even in New South Wales if Abbott remains Prime Minister until that election, especially not after what we have seen with preference flow changes in Queensland.

Postcount updates will flow from the top as I add them - apologies for the lack of detail early on; there's a fieldtrip and a decaying Prime Ministership competing for my attention.

10 comments:

  1. Kevin,
    Your scepticism about respondent-allocated preferences seems justified to me as a general principle. Intuitively, people don't have a recommendation from the Party of their primary choice, as they (usually) do when casting their vote, which influences their choices.
    However, the qualification is that where a big swing is in prospect, particularly one which is likely to see a change of Government, last election preferences seem likely to be a less accurate predictor. This was true to a marked extent in Queensland in 2012 and 2015. Even neutrals wanted Labor gone in 2012, just as neutrals* this time wanted the LNP severely shaken, if not thrown out. That means that some-one voting for a non-major Party* is likely to have a reversed choice of major Party preference in these successive elections.
    * They may not be precisely the same voters at the two elections.

    ReplyDelete
    Replies
    1. Agreed. It could be a function of swing as well as a function of primary vote.

      Delete
  2. I don't know if you'll get this, Kevin, but Lockyer seems to have sprung back to life, with Hanson managing to enter the 2PP lead with 5 out of 37 booths counted. I think that this is because the booths with a 2PP count have low Labor votes, and Labor voters will probably have stopped numbering at Hanson. However, the 'number every box and put the LNP last' message put out by the ALP might accidentally give Lockyer to Hanson.

    ReplyDelete
    Replies
    1. Ta. I'll try to keep an eye on this - at present William Bowe thinks it's going to blow over as more booths have their 2PPs counted.

      Delete
  3. Hey Kevin,

    Just wondering if you can comment on something I think a lot of the analysts have been missing today.

    Late polling and postals has been breaking heavily for the LNP, as they do, and "those in the know", like Antony Green, are insisting that Ferny Grove cannot be won and Whitsundays most likely won't be won by the LNP because of absentees and their tendency to support Labor. I mentioned to him earlier, that if he is to include Mansfield as a seat which is still 'in doubt', due to absentee's breaking for Labor, then the same must apply to Bundaberg (I didn't get the answer I was hoping for).

    The reason I even picked up on it today was when I saw how drastically the swing had changed since Saturday night. At one point it was registering a swing of 24% to the ALP. This morning it was at 21.2%. By lunch time it had been reduced to 20.1%.

    Since that time, it has been reduced to 19.6%, and a margin of just 646 votes. But this final drop was AFTER the ECQ had turned off the indicative 2pp count in the seat. The break of preferences in the seat coming from postals is dramatically different from those registered on polling day. The current calculator on the ABC site is using estimated preferences from results on Saturday. Granted, it's not anecdotal, but the drop in 2pp swing from the first batch of preferences added this morning compared to this afternoon was significantly different (they were both pretty similar sized, two batches of around 1,000).

    This made me think the seat was certainly in play. On further research, using the 2012 election results as a guide, there is only about 600 or so postals left. And 1,500 absentee votes.

    Link here: http://results.ecq.qld.gov.au/elections/state/State2012/results/booth11.html

    As you can see, the postal and absentee votes broke for the LNP much better than their polling day performance. Postals were 60-20 to the LNP, with slightly lower for everyone else.

    Absentee votes were 52-19, with higher green and FF

    Now 2015: http://results.ecq.qld.gov.au/elections/state/State2015/results/booth11.html

    Currently the postals have broken 54-31 to the LNP, with everyone else down. But importantly, only 6% down on their 2012 result (60 to 54).

    If a similar result is played out with the absentee votes, they will be gaining 46-35, which on 1,500 comes to 690-525, or 165 primaries. If the postals keep breaking as they are, 324-186 (138 primaries). Which suddenly cuts the margin to 300. It would be a fair assumption to say that preferences for the 'other' votes in the last 2,000 still to be counted will probably break for the LNP, given the large FF vote last time without a new home, and the large KAP/PUP vote that will exhaust. This is all without taking into account the batch of over 1,000 postals that are currently being accounted for with incorrect preference flows. Big gains on primaries also has a larger effect under OPV.

    I know, a lot of writing and assumptions to still return what is probably going to be an ALP retain, albeit very slim. It's frustrating not to know the impact of incorrect preference assumptions on the earlier 1,000 postal votes. One thing is certain, the margin is lower the 646 and it will be tighter then everyone is predicting. Poor decision to stop indicative 2pp preference counts in this seat, given the chance of a hung parliament.

    ReplyDelete
    Replies
    1. With increased prepoll voting and the circumstances of this campaign I am more cautious than Antony these days re applying the old wisdom about seats on slightly more than 1% being very unlikely to shift in post-counting. I've put a flag on Bundaberg in the summary at the top of the post and referenced your comment. I'm alas not yet in a position to look at seats closely but I agree that if the lead size in Bundaberg is similar to a number of the other seats sitting on about 1% it too should be treated with caution.

      Delete
    2. Cheers,

      Not much counting in Bundaberg today.

      The margin has been cut to 616, with over 2,000 votes remaining. Postals are officially splitting 61-39 to the LNP. They will continue to do well on the remaining 600 postals and pick up at least another 125 votes. Everything comes down to the 1,600 outstanding absentee votes. A similar strong showing amongst absentees relative to their performance in 2012 with consideration to the current swing against them, will see this come right down to the wire.

      More interestingly, late this afternoon the ECQ turned indicative 2pp counting back on in this seat, suggesting a close result.

      On Monday morning when the ECQ initially turned off indicative counting for all seats other than those that remained in doubt, 5 seats, Ferny Grove, Mansfield, Mt Ommaney, Maryborough and Whitsundays kept showing notional counts.

      In the afternoon, they added Gaven and Lockyer, and took off Ferny Grove and Mt Ommaney.

      Today, they took off Gaven and added Bundaberg and Mt Coo-tha.

      Not trying to sound like a dyed in the wool Liberal desperately clutching at straws (I disclose, I did vote for them) but it would appear they are expecting a legitimately close result in Mt Coo-tha as well.

      Of course, the same applies to Ferny Grove, where they are clearly expecting absentees to favour Labor. I agree with this, and expect Bundaberg (and Mt Coo-tha, although I am yet to look into this) to be much better prospects for the LNP. Clearly Whitsundays and Maryborough remain in doubt, with Mansfield in the same category as Bundaberg and Mt Coo-tha.

      Delete
    3. Ok, I've worked out why. Saxon Rice has done unbelievably well on postal votes in Mt Coo-tha. She's gained over 52.4% on the first 1,900 added. 10% higher than her booth primary and astonishingly up from the 52.14% she received in 2012.

      I'm not getting too excited yet, the current margin is 1,148 (yet to include the 2,000 postals, which presumably will be counted for a notional 2pp today).

      She has gained over 400 primaries on Labor. But this is the seat with the highest Greens vote in the state, bar none.

      The booth results on the day produced 42.8% LNP, 32.83% ALP and 22.3% Green.

      Under normal circumstances with OPV, this would be an easy LNP victory. However so far on the 20,000 notional 2pp counted, the preference flows to Labor have been 75+%. It's almost compulsory preferential like but without the leakage to the LNP.

      In 2012 on postals, the LNP was 52.14% (+3.5%), Labor 29.5% (+1), Greens 15.2% (-5).

      In 2015 the LNP is 52.37% (+10%), Labor 31.8%(-1%), Greens 13.7% (-9%)

      On absentee's in 2012 LNP were 45.7% (-1.5%), Labor 24% (-4%), Greens 25 (+5%). The important thing to note here is that Greens actually outpolled Labor on absentee's.

      The basic point here is obvious. Given the LNP will gain on remaining postals and have its primary hold up on absentee's while Labor declines, the future of the seat is now entirely dependent on the enormous greens preference flows holding up.

      If we work on the assumption the LNP maintains 40%, Labor 28% and Greens 27% on the 3,300 outstanding absentee's, LNP 1320, ALP 924, Greens 891.

      The LNP will have gained almost 1,000 first preference votes on Labor and closed the gap, but once Green preferences begin breaking for Labor it will be an easy retain.

      This will only get interesting if Rice's primary vote on absentee's holds up similar to the last election, like it has done on postals (at 43 or 44%, given her primary vote is already 43.6 and will need to improve further). If this happens, the Labor vote will be below the Greens vote on absentees and preferences become crucial. Anything like the green preference flows we saw on Saturday and she would be toast regardless.

      Put in the pile of 'potentially interesting' rather than in doubt. Rice has done well in a traditional Labor electorate, with the highest Greens vote in the state and a 5.6% margin to even be in the race. Ironically if she we're to miraculously hold on she'd be a certainty for the ministry.

      Delete
  4. I've found something extremely interesting in Pumicestone this morning...

    While scrolling through the booth results, it appears the electoral commission have miss apportioned a batch of 940 votes incorrectly to an independent polling 1%.

    Booth no. 3, Bellara, Lisa France has 23 votes (1%) in a booth she polled 1,400 (55%) in 2012.

    http://results.ecq.qld.gov.au/elections/state/State2015/results/booth68.html

    If you add this batch to her tally, her primary vote increases from 38.8% to 43%.

    A quick look at the ABC website, and the current 2pp has an absolutely ridiculous margin of 2,100.

    Starting from correct primary vote bases of LNP 12,864 and ALP 12,235. If you apply the 45% ALP, 15% LNP, 39% exhaust test it leaves the 2pp preferred margin at 13,591-14,418, or an 827 vote margin to the ALP.

    Clearly this is still ludicrously inaccurate, given that the that the Palmer vote is 600 votes higher than the Greens, and almost 1,200 votes are split between two independents running in a seat which was at one stage held by One Nation.

    To make my point, If I use preference assumptions from the last election in this seat: Greens vote, 47% Exhaust, 78-22 to ALP the rest and KAP (PUP for this exercise) 48% exhaust, 55-45 to the LNP. And assume the Independents neutral, Lisa France would be leading by over 200 votes.

    The real story is more likely to be the ALP leading by around 100-200. With absentee's it will likely blow out further. However this is still close enough to warrant attention.

    I've sent the electoral commission an email, but I have no idea what the appropriate channel for sending an inquiry like this. Any suggests would be appreciated.

    ReplyDelete
  5. I note that the site now has primaries at 13010 to 12379 and reasonable booth figures so the glitch seems to have been fixed. On this there is a 2PP margin of 854, about per your estimate based on a 45-15-39 split. A possible reason the normal 2015 flow of preferences is roughly holding up despite a lowish Green vote or any residual ONP tendencies is that the share of non-Green minor party votes that is held by PUP is much higher (64%) than in the state as a whole (40%). I'd expect that PUP preferences have a better flow to Labor than those of KAP and other non-Green minors overall.

    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.