Queensland: The Election Polls Forgot

Queensland Aggregate: Labor 35.9 LNP 33.9 One Nation 13.3 Green 9.6 Other 7.3
2PP off this aggregate: 52.5 to Labor
Seat projection off this aggregate: Labor 51 LNP 37 PHON 3 KAP 2 
(Greens may win 1-2 seats instead, but not enough reliable evidence re Greens' chances)
Uncertainties: limited polling dominated by a single pollster, change in preferencing method and likely preferencing behaviour

Nearly three years ago an incumbent Queensland government was leading about 52-48 in the polls and polling analysts generally though that it might lose but would probably be returned.  We were wrong; it did very narrowly lose, mainly because of a radical shift in preferencing behaviour by voters for minor parties.

Three years later, some of the problems are the same.  A radical shift in preferencing behaviour is possible in theory, because Pauline Hanson's One Nation was just a minor presence last election but is now gathering a double-digit vote.  Moreover, some polls have showed One Nation voters as much more likely to preference the conservative side of politics than they used to be - a situation consistent with the party attracting a lot of ex-Coalition supporters.



There is a new problem in terms of how much we can trust polling data.  Nearly all of the polling data for this election has come from a single firm, albeit one (the Newspoll/Galaxy stable - see final Newspoll here) with an excellent record at recent elections.  The main challenger, ReachTEL, has repeated its form from the WA election where its results were weaker for the government than Newspoll's, until it moved into line with it very late in the piece.  Apart from that, there were some pretty old monthly Essentials but that's about all that's been seen.  We're left needing a lot of faith - however justified by the past evidence - in a single pollster.   This is quite odd coming after a same-sex marriage postal vote that attracted interest from at least ten different pollsters - and perhaps that is the explanation there.

In the big picture, should what we are seeing in the polling (a small move to Labor, concentrated in the state's south-east) surprise?  Historically, first term state governments generally win, with some exceptions.  State governments generally also win when the opposite party is in power federally, again with rare exceptions.  Moreover, state governments of the latter kind tend to gain seats.  First term state governments that are of the opposite party to the federal government?  In modern times they always win, and they sometimes go from very tenuous positions to big majorities.

For all that, the Palaszczuk government comes with a few warning signs.  Voters have generally forgiven governments that come to office in minority, but this government slipped further in minority through its term with losses to the crossbench that at times made it hard to control the parliament.  New state premiers who are able to indulge in Canberra-bashing have often been very popular, but these are hard times to be a popular leader in for long, and the Premier's net rating as she goes to the election is a mediocre -7 (40% satisfied, 47% dissatisfied).  The government has also been indecisive on the major issues associated with the Adani coal mine, frequently seeking to have a foot in both the pro- and anti- camps, at the risk of impressing no-one.

I don't have much detail to add to my previous polling and modelling comments, but just a few points to add:

* Tim Nicholls is very unpopular.  His -27 netsat (27% satisfied, 54% dissatisfied) is seven points worse than the worst ever for a winning opposition leader (Jeff Kennett).  It's true that the link between opposition leader ratings and voting intention hardly exists, but there have nonetheless been only two state winners from opposition with ratings worse than -5 (Kennett and Richard Court).

* Both Galaxy and Newspoll latest releases only put One Nation on the readout in the seats they are contesting.  The six-point drop in One Nation support in the final Galaxy was partly for this reason and partly because the pollster appears to have registered a drop in support regardless.  (William Bowe makes the drop about three points, so I've adopted this.)

* My mucking about with Alex Jago's calculator suggests One Nation can still win a few seats, though it's possible they will wipe out entirely.  An important tweak here is to add the above-mentioned three points back in.   For the time being the modelling says they have good chances in Lockyer, Mirani and Hinchinbrook - all seats that have had no public polling, and all seats where they should be able to push the LNP into third (which is much easier for them than if they have Labor there.)  Perhaps they won't win those and may win others, but they face a range of problems in other seats where they are being given chances to win.

* I am giving KAP their two seats, in the complete absence of useful evidence on the question, but there's some thought out there that they could also pick up Warrego.  There are also a few indies about who have attracted some attention, but no polls suggesting any are close to winning.

I've tried a few different modelling setups to cater for One Nation's bizarre preferencing behaviour of mostly preferencing against the "incumbents".  In the quieter version, the One Nation preference decisions are mostly not enough to unsettle the advantages in personal vote and required swing that the parties hold in various seats.  On a 2PP basis in this version, the model says Labor should gain Redlands and Glass House, that Whitsunday is a tossup, and that Labor have realistic chances - not in order of chance - in: Bonney, Everton, Gaven, Maiwar, Hinchinbrook (if PHON don't get it), Chatsworth, Nicklin (!), Caloundra and Toowoomba North.  On the LNP side, Bundaberg and Maryborough are tossup gains according to the model, while realistic chances also exist in Mansfield, Mt Ommaney and Burdekin (all Labor notionals held by the LNP), Keppel, Mirani (if PHON don't get it) and a seat which has had very little attention despite its incumbent being given the boot, Pumicestone.

If we look at seat polls for this list, the balance is slightly more friendly to the LNP than the assessment suggests, but not to a point that anything much can be read into it.

In the "mindless carnage" version, I assumed that One Nation preference decisions were directing around a quarter of their voters, causing incumbents on both sides to be smashed.  What I find in this version is that Labor's realistic losses seem to be capped around six or seven seats with an average of four, because above a certain level either the margins are too high or One Nation isn't much of a factor in those electorates.  On the other hand, the LNP's losses in this case could run well up the pendulum with an expected total of 12 losses, so Labor actually does better in this case.  I am sceptical of this scenario but note it just in case!

How would Labor lose?  Well, the first possibility is that the polls just all have the primary vote plain wrong (especially, perhaps, an underestimate of One Nation); it's got to happen some day and it wouldn't have to be by much in this case to knock that 52.5 2PP down to something more loseable.  If the Labor 2PP is really only, say, 50.5 then we'll probably get a hung parliament.

The second is that One Nation voters might actually do something sensible.  They might reason that on current polling there's a real chance Labor will win a majority while the LNP are not likely to.  Presumably they would want their party to get the balance of power, so the best way to do that would be preference the LNP strongly.  If One Nation voters decided to preference the LNP three to one everywhere and ignored whatever the how to vote card said, then Labor would be looking very dicey.  However, this doesn't seem likely.

A final factor I should mention is that the informal vote is likely to increase thanks to the reintroduction of compulsory preferencing.  Possibly Labor and One Nation will both be hit by this more than other parties.

One point I do want to make that I think has been ignored is this: everyone is talking about One Nation preferences but Green preferences are being more or less ignored.  At the 2015 election, under optional preferencing, Labor gained at around .5 votes/preference from Green preferences.  That's the equivalent of a 75:25 split.  But if that split becomes 80:20, that's an extra .1 of a point to Labor per point of Greens preferences.  That's already factored in to the statewide picture.  However, in particular seats where the Green vote is high, it is not.  There may be inner city seats (especially Maiwar, should the Greens fail to get it for themselves) that are closer on a 2PP basis than they look because of this.  As jaded as Greens voters may well be with Labor's constant switching on Adani, another (albeit unglamorous) green issue, land clearing, has shown some more serious distance between the major parties.

So the polling evidence favours a modest Labor retain, but it's close enough and the evidence is thin enough that it's not even near a sure thing.  On the other hand, a repeat of the WA blowout leading to an easy Labor win is also conceivable.  I'll be back with live commentary tonight.

PS Essential took a poll over the past month with estimated 53-47 to Labor off primaries of Labor 36, LNP 32, PHON 15, Greens 10, KAP 1, Others 6. Essential's figure for One Nation is statewide, with One Nation included for seats in which it isn't running, so that continues a pattern of Essential having lowish readings for One Nation compared to others.