Friday, October 30, 2020

Queensland 2020: Rolling Final Days Roundup

I'll be covering the Queensland election here on election night and in detail throughout the postcount, though on some days I may be busy with other things during the day.  

Another Queensland election count is a day and a bit away and there's a distinct lack of data for the usual polling aggregation/modelling type game, so I've decided to start a rolling roundup article which will cover a number of possible themes.  When new polls appear - assuming they do - new sections covering them will be posted at the top of the article. I also have a new article in The Guardian.  I should clarify that I didn't speak to any of the nameless insiders personally, and was going off other media reports.  There were reports since I filed that one that unnamed Labor insiders were more confident yesterday because internal polls had supposedly shown an uptick following one of those debates that nobody much watches.  This narrative was gleefully and even gullibly snapped up by Sky, perhaps because it was useful for attacking Gladys Berejiklian.  The NSW Premier will doubtless be scapegoated in the event of an LNP loss even if there is nothing special in the early/late voting history to justify such a charge.


Final Newspoll

The final Newspoll gives Labor a 51.5-48.5 lead off primaries of Labor 37 LNP 36 Greens 11 One Nation 10 leaving 6 for the rest (I predict it will be more!).  The leader ratings are also largely unchanged: Palaszczuk 62-33, Frecklington 35-48 and Palaszczuk leads 56-30 (see older ratings below).

For the record this is my seat model's attempt to project what a 51.5% statewide 2PP as adjusted for regional swings in the first YouGov poll might look like (I have also added a small volume of seat poll juice to the formulas):

(Click for larger version).  The model gives an estimate of the probability of the current holding side winning the 2PP - not necessarily the seat in cases like McConnel, Oodgeroo, and any ON/KAP threat seat - on the assumption that the 2PP is 51.5.  This is not a prediction that this will happen, because the polls might be wrong.  The model will also be blind in particular seats, and some of these probabilities could therefore be silly.  For instance it currently treats Bundaberg as regional, but there's a view that Bundaberg should be treated as an extension of the Gold/Sunshine Coasts, and Labor might then win it back.  The projection for 2PP wins in seats not won by crossbenchers in 2017 comes out to ALP 49 LNP 38.  Labor being likely to lose South Brisbane to the Greens is also a projection well backed by objective data (whether it happens or not) so that takes Labor back to 48 seats.  They could lose to other crossbench challenges too (eg McConnel, KAP/ON seats) but this is also a risk for the LNP.  It's extremely difficult to get the LNP to a majority off the Newspoll 2PP, so they need Newspoll to be wrong - even an LNP minority would need about an extra 5-6 2PP wins over the projection. 

If the Greens and KAP (or One Nation) both fire it becomes less likely Labor would win a majority off this 2PP, but off a slightly higher one they still might.  

It's possible that a status-quo 2PP (if it happens) see only a handful of seat changes as in 2017, but it could see more variation in seat swings and more seats changing hands in both directions.  

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

(earlier text below)

Polls: Too Few, Too Old

As I write there have been only three statewide public polls (a YouGov, a Newspoll and a Morgan SMS) and the youngest data in these is already two weeks old.  The statewide polling picture all those days ago was next to zero swing - maybe a fractional swing to Labor.  It's well within the range of recent Australian state polling errors that that any final aggregated estimate could be wrong by a couple of points either way; it's plausible (but less likely) it could be wrong by more than 3 points, a la Victoria 2018.  We don't have a recent precedent for a 4 point 2PP error, but those aren't impossible either.  The current YouGov polls are experimental (but it seems like a promising experiment) and the record of Morgan SMS is not that flash.  Let's see if we get a stronger signal between now and polling day, but at least for Labor, for now, there's no evidence of a statewide swing against it.

All regional data except Morgan's (which is obviously nonsense, since it has the Greens primary and the Labor 2PP higher in regional Queensland) backs in the ideas that:

* Labor could lose seats in the north

* Labor could gain seats in the Gold and Sunshine Coasts, where it currently holds very little

* There might be some swing to the LNP in Brisbane, but it probably won't do much if there is

* The One Nation vote is down, but that might be a little or it might be a lot

* Labor could lose seats in Brisbane to the Greens

Nearly all the public evidence supporting these claims comes from one pollster.  If the LNP holds the coasts and the inner city and gains seats in the north, it can win.  If Labor holds the north and gains seats elsewhere, it can win comfortably.  Or something completely different might happen.  

See this previous article for modelling on the basis that the picture in statewide polling so far is broadly correct - which of course may not be the case.

Labor Is Killing It On Non-Vote Polling

Despite the close voting intention polls, Labor is doing extremely well on poll questions other than voting intentions.  Recent leadership readings have included:

Palaszczuk Satisfaction/Approval: 63-33 (Newspoll), 62-28 (Essential), 57-27 (YouGov), 63-33 (earlier Newspoll)

Frecklington Satisfaction/Approval: 37-44 (Newspoll), 29-32 (YouGov)

Preferred/Better Premier (skews to incumbents, for Prime Ministers by c. 15 points) Palaszczuk leads 56-32 (Newspoll), 48-22 (YouGov)

Better at managing economic recovery and jobs (exact question and answer options not seen): Palaszczuk/Labor leads 40-27 (YouGov)

Better to manage economy: Palaszczuk leads 50-26 (Ipsos)

A disclaimer should be noted with the Ipsos figure - this is an online Ipsos poll format that is untested for voting intentions at any election.  However if you told me only these numbers and not the voting intentions - I would say the election was almost certainly already over.  And I'd only add the "almost" because of one historic Queensland exception ...

The Ghosts Of Goss and Churchill

In 1995 Labor under Wayne Goss initially won the election by one seat, but had their win in Mundingburra overturned in the courts, then lost the 1996 by-election and office (albeit not for long). 

During the campaign, Goss had led Opposition Leader Rob Borbidge in Newspoll 61-22 then 57-24, and his personal ratings were high (albeit slipping at the end), 60-32 then 54-38.  Borbidge's ratings were nothing special (his final rating was slightly negative, 37-42).  But the Coalition was competitive on primary votes, with the last two Newspolls getting 45% for Labor to 45.5% for the combined Coalition parties.

On election day the Labor vote was 2.1 points lower than the Newspolls had it, and the Coalition vote 3.5 votes higher.  Labor lost the 2PP so heavily (53.3% to Coalition) that they were lucky to even make it as far as Mundingburra.  So there is a precedent (albeit a lonely one, hospital-pass Premiers excluded) for a Premier to not exactly win an election while being very popular.  But it only goes so far, because of federal drag (Labor was in office federally).  

Another argument that's reportedly being run (mainly by Labor insiders afflicted by either pre-election nerves or pre-emptive expectation management) is that Queensland 2020 is a bit like UK 1945.  That is, Labor has won the war with COVID and won it so well and for so long that nobody still cares about it, voters think the state is safe whatever happens, and they look to who will run the state best in what they expect to be a COVID-free future whoever is in government.  In this regard, the LNP project more enthusiasm and excitement (even if some of it is about schemes that are probably daft and that won't actually be funded beyond study level in their first term anyway) and have also made a play for the hip pocket with their $300 car rego rebate scheme.  And pre-COVID, Labor was showing quite a lot of wear and tear.  

One Nation Preferences

At the last election, One Nation preferenced the LNP in considerably more seats than Labor, though annoyingly I can't relocate any records of exactly which ones they were.  At this election One Nation is not preferencing either major party (despite getting LNP preferences on a platter) and is instead issuing double-sided preference cards; Katters Australian Party is doing the same.  Notably on One Nation's website the version with the LNP candidate ahead is given first in all bar three seats, the exceptions being Mundingburra (where the LNP candidate has had social media issues), Stretton (where the LNP candidate has been accused, not entirely accurately, of excessive pro-China propaganda), and Scenic Rim (not sure what the background is there).  

Any shifting in One Nation preferences could harm Labor in seats where it benefited from being preferenced by One Nation last time, most notably Aspley and Redlands. But the two-sided card could help Labor in seats where it did really badly on ON preferences last time, including Townsville, Mansfield, Barron River, Cairns and Pumicestone.  This would have less than 1% impact on the 2PP per seat in most cases, but some of these seats could be very close.

Black Swan Watch: The KAPslide!

 I will be watching the performance and preferences of the Katters Australian Party on election night with a lot of interest.  In 2017 KAP upstaged One Nation by winning three seats.  The seats it holds are Traeger, Hill and Hinchinbrook, and they together cover pretty much the footprint of Bob Katter's federal seat of Kennedy.  Nick Dametto won Hinchinbrook by beating Labor into third by 2%, using Labor preferences to overtake One Nation and then using One Nation preferences to beat the LNP. The last two steps were relatively comfortable so a crash in the One Nation vote wouldn't necessarily cause Dametto any problems.  The question is whether this winning from third model might be replicable in other seats, especially if the One Nation vote declines but the KAP vote doesn't.  Some to watch are:

* Cook, where KAP missed third place by 0.5% last time; probably Cynthia Liu picks up a personal vote here but if there is a swing against Labor counteracting it then winning from third might be doable.

* Mundingburra, where KAP missed third by 2.35% last time and nobody was near 50% (that said, the Newspoll suggests polarisation towards the majors that would give KAP no chance).

* Thuringowa, where KAP missed third by 3.6% last time and the highest primary vote was a feeble 32.2%.  The party almost won this seat in 2012.

* Burdekin, which neighbours Kennedy and where primary votes were split quite evenly between Labor, the LNP and One Nation in 2017 with KAP not running.  The KAP candidate is Sam Cox, the 2012-2015 LNP MP for Thuringowa.  

* Whitsunday, where they are the second favourite in seat betting though I can't quite see the logic here, since they only got 9.5% last time.  With the former LNP member Jason Costigan now running with his own party and nobody exceeding 32.2% last time, this seat could be very messy.

It may be that polarisation towards major parties means nothing happens here, but if that polarisation doesn't happen to the extent expected and votes shift from One Nation to KAP, watch out!  There's potential for the party to win lots of seats from 2-3% of the state vote.  

Oodgeroo

Almost any election these days seems to have this script somewhere - the local high-profile moderate independent (socially liberal and environmentally concerned) who threatens the otherwise safe conservative - think Zali Steggall in Warringah, but don't forget this plan was tried in many other seats around the nation and Warringah was the only one it worked in.  In this case there's an added incentive via a proposed port development that is apparently much hated - which has the incumbent, the LNP's Mark Robinson, suddenly scurring to find some environmental cred.  Oodgeroo is a good target for an indie because there are reasonably large Green and Labor votes to serve as stepladders from third place on primaries if necessary (though Claire Richardson might well outpoll Labor here) and the LNP vote only needs to be knocked down from 52.4% to, say, 44% and game on.  There have been reports that some kind of poll for Richardson's campaign has her ahead 51-49 but no details of methods, sample size, pollster etc to scrutinise (some reference to an "exit poll" that may or may not be the same finding); in any case commissioned polls tend to favour their sponsors.  There have also been reports that the LNP think it is under control.  

There is some thought that the pandemic situation makes it hard for independents, but I'm not sure that is true - a vaguely comparable indie almost won the Tasmanian Legislative Council seat of Rosevears against a rampantly popular state government, for example, though there were plenty of factors in the indie's favour there too.  What does create difficulty is that the performance of indies on postals and prepolls tends to be poor, and there will be a lot of those vote types in this case.  

There has been less attention regarding Margaret Keech in Macalister, a former Labor MP attempting to unseat Labor in a seat where an independent was also a threat in 2017.

Die GrĂ¼nen

Everybody knows the Greens are the world champions (among parties with any actual support, anyway) when it comes to chalking up their wins before the event. Unfortunately a spectacular result in the ACT has done nothing to deter this tendency and their supporters have recently got bored with declaring incipient victories in South Brisbane and McConnel, and set their sights on Cooper.  Cooper is vacant following the retirement of Kate Jones who will take a considerable personal vote with her, but the Greens start 15% behind the LNP and 20% behind Labor, so this seems very ambitious.  A Newspoll seat poll suggested Jackie Trad is in massive trouble in South Brisbane (but it's a seat poll, and via experimental methods at that, so not conclusive) but there has been no fresh public evidence re McConnel or Maiwar.  Trad fits the profile of a troubled incumbent (the sort most likely to lose to crossbenchers at federal elections) but at state level an incumbent need not be troubled to lose to the Greens; they can just be a victim of shifting demographics and inner-city intensification of the Green vote.  

Betting

Betting is not necessarily predictive and indeed is often completely wrong, but I like to keep an eye on it to see how well or badly it actually does.  On the headline, Labor has been firming rapidly in recent weeks after taking a very long time to be favourite at all, and is now favourite by 1.20 vs 4.00, an implied 77% chance of winning.   Labor is also given an implied 57% chance of winning in its own right.  The former would be overconfident if based on polling alone, but not if based on historic probabilities.  If Labor do win outright, the question about these markets will be what took them so long (for the second election in a row).

The seat betting has quite a guarded view of Labor's chances.  Incumbents are favourite in all seats except:

* Barron River, Keppel and Townsville, expected to switch from Labor to LNP.
* Pumicestone and Currumbin, expected to switch from LNP to Labor
* McConnel and South Brisbane, expected to switch from Labor to Greens
* Burdekin, an LNP seat that is a tie between LNP and KAP
* Whitsunday, the North Queensland First seat (Costigan) where the LNP are favourites but at longer than $2

(Labor are also favourites but at longer than $2 in their own seat of Thuringowa)

If all these played out, Labor would be put into minority but would be able to choose between the Greens and KAP, primarily, for support, or to operate as a minority government on the assumption that the Greens will never put the LNP in.  There isn't much of an imbalance in seats that the markets have serious doubt about (one, incidentally, is Oodgeroo where Richardson is very nearly favourite), so there seems to be some discrepancy between this seat betting and the majority government odds.  

Election Night

I hope election night won't be a repeat of the shambles of the by-election web displays earlier this year.  Things to watch will include a vote mix with prepolls in the high 30s as a percentage of turnout, postals probably in the low 20s, on the day voting perhaps in the low 30s.  It will be difficult to project seats early in the night and some seats will have high out-of-electorate prepoll rates that may require counting into subsequent days before too much is known about them. Differences between on-the-day voting and postals may weaken compared to the past (especially as the postal count goes on) and we should be careful especially of projections that don't include actual postal counts.  

No comments:

Post a Comment