By any normal measure the Colin Barnett Liberal government in Western Australia is in its final weeks. A weak economy, long incumbency, federal drag (and from an unpopular federal government at that), disunity, a disliked Premier and bad 2PP polling would together spell "GAME OVER". It's actually not easy to find electoral precedents for a government with quite so many things against it all at once.
The close nature of some recent 2PP polling, however, has given some (but not all) Liberals serious hope that they can sandbag their way back into power, especially if they can get One Nation preferences. The latest Newspoll has, for now, pretty much pulled that rug out from underneath their feet. It's much too early to say that it's all over, but things do not look good.
The Newspoll, taken this week, has a 2PP vote of 54:46 to Labor with primaries of Liberal 30 (down 4 since November) National 5 (down 1) Labor 38 (down 3) Green 9 One Nation 13 (up 10) and remaining others 5 (down 2). As suggested by William Bowe in some essential reading on this contest, it is plausible some change in treatment in the "readout" has contributed to the polled One Nation surge and that their support level back in November was much higher than 3%. This would certainly be consistent with One Nation's federal vote, which has moved from about 6% to 9-10% in the same time.
If we compare this to the May Newspoll which also had a 54:46 2PP, the 13% credited to One Nation has come at the expense of: five points from the Liberals and Nationals, four from Labor, two from the Greens and two from miscellaneous "others". This suggests the 2PP picture in this Newspoll is likely on average to be around 54:46 exactly, as opposed to the May Newspoll which, as I previously showed, was likely on average to have been 53-point-something.
Reports of the Newspoll suggest that One Nation preferences are splitting evenly, though I have not yet seen it made explicitly clear that this was canvassed rather than being assumed. In any case it is a moot assessment should One Nation consistently distribute preferences to one party. But they say they probably won't do that, and as William points out again in this radio interview as well as his post, One Nation preferences are not very directable. One Nation does not yet have the ground game for how-to-vote handouts on the scale of the majors, and even if it did, its supporters tend to make up their own minds. The suggestion is that how-to-vote preferences from One Nation are only worth about a point per seat on average. More than nothing, but not a game-changer from 46:54 down, especially not if Green preferences flow more strongly to Labor than in the past.
Some more evidence on the directability of One Nation preferences (or not) comes from the 1998 federal election, the party's best House of Representatives performance to date. The overall One Nation split was 53.7% to Coalition and the party's preferencing strategy was infamously all over the shop, often targeting incumbents from both parties. Yet no seat nationwide where One Nation made the final three (which was quite common) saw a split of One Nation preferences (albeit including preferences received by One Nation from other parties) outside the range of 33.8% to 68.1% to Coalition. Also when One Nation preferences did split more than 60-40 either way it was normally in safe seats (nearly always in the winner's favour.)
One important takeaway from 1998 is that when One Nation preferences did flow strongly to one major party in a close seat, it was usually to the Nationals. Indeed the six strongest two-candidate ON flows nationwide were to Nationals. (It is likely Brendan Nelson (LIB) in ultra-safe Bradfield would have been in this mix had the ON candidate made the final two there though.) Now this is not good news for the Barnett team at all because the seats held by the Nationals are way up the pendulum and generally not winnable for Labor, even had they not been bolstered by the Royalties for Regions program. So a strong preference flow to the WA Nationals in their seven seats (and others they're competitive in) will just mean less preferences to the Government everywhere else, for a given preference flow. We might even see One Nation flow slightly to Labor in seats that Labor is targeting. If One Nation even preferences Labor in some of Labor's more ambitious targets, that could be bad for Barnett.
So we have to see what One Nation does, but nothing they do with preferences would be at all likely to save the Liberals off their primary vote polled in this Newspoll.
The Newspoll shows some improvement for Barnett in personal ratings, with his net satisfaction up eight points to -25 (32-57) and Mark McGowan's better Premier lead cut to 12 points (44-32). These are still however dire figures for the incumbent.
The following graph based on Newspoll data immediately before 51 state elections shows where Barnett is on leadership scores in relation to past winners and losers. In the case of some three-cornered Tasmanian and Queensland better Premier polls I have applied adjustments to get a two-way estimate for the major parties:
This is a pretty simple story. If a Premier's netsat on election eve is positive they usually win, and if it is negative they usually lose. Better Premier has a house edge to incumbents, and there the break-even point is a lead of about eleven points. If we simply add the two together, those with a combined score of +11 or better have an 83% success rate (I'm counting Queensland 1995 as a half) and those below +11 have saluted just 26% of the time.
But Barnett isn't just below +11, he is way, way, way below it. His combined score of -37 is 23 points below the worst of the winners (Tasmania's David Bartlett in 2010, who lost his majority), and exceeds only two of the losers (Lara Giddings and Anna Bligh, whose governments were thrashed). Normally Premiers who are this much on the nose do not make it to an election unless the government is so clearly doomed that changing leaders is pointless.
One model I did off these data put the chance of a Barnett eviction - should his scores not improve from here - at a whopping 89%. There aren't a lot of data points near him, so perhaps that's a bit overconfident, but it can't be so by that much. Of course, the usual disclaimers, every election is different, and records are there to be broken. Plus, this is the first test of the new Newspoll at any state election, after a brilliant performance in the federal poll.
As for Mark McGowan, his netsat of +12 (46-34) has hardly moved an inch from last time, when it was +13 (46-33). Steady as she goes then for the perceived Premier in waiting.
There was also a ReachTEL taken in mid-January which had the Liberals on 36.1, Nationals 6, Labor 35, Green 6.7, One Nation 10.8, Others 4.5. The published 2PP was 52-48 to Labor, which must have been by respondent preferences, as on last-election preferences this would come out to about 50-50 or maybe even a very small Coalition lead. So it may well be that ReachTEL's One Nation supporters were Labor-leaning, or their Green supporters very Labor-leading. Respondent-allocated preferences not only tend to skew to Labor, but are also volatile from poll to poll. However, we shouldn't ignore them entirely as evidence, because we are looking at a very large primary vote swing, which might well come with a fair preferencing shift in Labor's favour.
All up this one's a very different beast to Newspoll. If I just had this poll and knew nothing else I'd back the government to win. However that hasn't been the trend of other polling, and the Newspoll at least cancels it out.
There wasn't much else reported except that Mark McGowan led Barnett on a preferred-premier forced choice 55.7-44.3 and Barnett also trailed his deputy Liza Harvey 46.1-53.9. Normally Newspoll shows a house edge to incumbents that ReachTEL doesn't, which can be interpreted as voters who dislike both leaders but slightly prefer the Opposition Leader opting for the neutral option. While other interpretations exist, in this case I suspect that very few voters actually do dislike both leaders in this campaign. McGowan seems to be quite inoffensive and probably those disapproving of him in the Newspoll are Liberal/National supporters, who are also about the only voters left who like Barnett.
The only other statewide public poll we've had is a 51:49 to Labor Essential sample taken over three months, which I'd regard as very old rope now. Concerning internal polls, Labor are spruiking a 58-42 in their favour in Joondalup (Lib, 10.4%), which I'd wager is a cherrypicked result, but probably still one the Liberals should be worried about even so. One Nation's Colin Tincknell has some interesting internal claims from ten seats, but we don't know who the pollster was at this stage so we can't say how reliable they are. (At least they can't be worse than PUP's famous so-called "internal polling").
Translating The 2PP To Seats
As usual Antony Green has a calculator up which you can use to translate various poll 2PPs to seats assuming uniform swing. Over here, however, we never assume uniform swing, for the reason that it never happens. Also, since sophomore effect is both a theory and a fact, I think the "factor retiring and sophomore MPs" box ought to be locked on and you should have to fill out a stat dec declaring you identify as an election nut to turn the damn thing off.
Entering a 52:48 to Labor into the calculator gives the Coalition a 30-29 seat win even if you click the box. My randomly variable swing model gives the median result for 52:48 as a 31-28 seat win to ALP. The reason is that the uniform swing model would see the Coalition holding eight seats on 2% or less and another six on 2-4% compared to just two on 2% or less for Labor. In practice because seat swings are variable some of those eight Coalition seats would drop.
For 54:46 Antony's calculator gives 34-25 to Labor without ticking the box and 36-23 if you do. My randomly variable swing model falls in the middle, giving 35-24. A swing of this size pushes the Liberal sophomore-effect seats (those won from Labor last time) so far down in the mix that the sophomore bonus is wasted; meanwhile, the average swing in other seats will be a bit greater to adjust for that.
The worse news for the Liberals is that the swing won't be even randomly variable. It looks like they will be hit by swings above the average in the outer suburbs and probably lose more there than they can sandbag from below the waterline. So the result I'd get by merging these two polls with Newspoll weighted slightly higher (33-34 seats to Labor) could even be a conservative read of where things stand at the moment.
PHON Is The Black Swan
The exciting scenarios in this one all arise from the One Nation boom. Off 11-13% the party wouldn't have much impact on the outcome, and might win the odd seat here and there or mess with the exclusion order in some Liberal-vs-National seats.
What gets messy is if the PHON vote is actually being underestimated by the polls at present and then blows up still more by election day. Say, the polls are 55:45 a few weeks out and voters for both parties figure that the election is over so why not send the big parties a rocket. If PHON get, say, 20%, then they will make the top two in several seats and might take seats from either party.
The saving scenario for Barnett is that PHON do so well that they take away the seats Labor needs to get over the line and an unstable Liberal-National-One Nation government forms with a small majority. This doesn't seem that plausible to me. I think that if PHON are polling enough to take seats from Labor it will be because the Liberal primary vote has melted in the sun. In that case two or three seats at Labor's expense won't matter. PHON will be a bigger force in the Upper House where they are very likely to win some seats. But in times when politics as we knew it seems to be upending itself on a daily basis, we need to keep an eye out for any snowballing of support towards Australia's number one retro political party.
That's all for now but I expect to write more through the campaign. There won't be any live coverage (clashing commitment all that weekend) but I'll do post-commentary starting a few days after.
Privatisation Poll PS
Newspoll also did a poll on the Barnett government's plan to sell a 51% share of Western Power poles and wires in order to retire debt and fund schools. Liberal and National supporters were evenly split on the plan and it got a massive raspberry from everyone else, coming out with only 25% in favour and 61% against.
Privatisation polling is frequently misleading because voters are asked what they think about flogging stuff off without being told what will be done with the money. Seen one such poll with a 30-60ish result, you've seen them all. It's an instinctive "yuck" generator. When voters are told there is actually some point to the sale, the polls can get very different results, as shown with polling for the Baird government's similar plan at the 2015 NSW State election.
But in this case the poll question did tell voters what the purpose of the sale was, at least loosely, and the result was still abysmal. What it would have been had Newspoll not done that - it doesn't bear thinking about! We see a government with very little support for its plans outside the ranks of its own followers. If you're thinking it will get good preference flows, I wouldn't hold your breath.