If 2016 taught us anything with Brexit and the POTUS elections it is that polling cannot be trusted. Unfortunately I have neither the time nor creativity to work out another option before the weekend, so we will be using the tried-and-sometimes-true methods that rely on opinion polling. But first, the pendulum.
Divided along Coalition-ALP lines, the pendulum looks something identical to this:
|North West Central||NAT||11.5%|
However, it is important to note that six National Party seats (Central Wheatbelt, Kalgoorlie, Moore, North West Central, Roe and Warren-Blackwood) and one Liberal seat (Geraldton) are more closely contested within the Coalition than with the ALP:
|North West Central||NAT||10.5%|
This is a reasonably normal pattern in strong National-voting seats, since the two parties are similarly aligned and the base vote is enough to overcome the "spoiler effect" you would have in a first-past-the-post system (also aided by the absence of such a system). In all of these seats, the ALP is more than 10% behind the leading Coalition party, and in most cases more than 20%. It is usually safe to rule these seats as a Coalition win, but the swing in this election could be substantial enough to bring this into question and WA has the weakest LIB-NAT Coalition in the country (to the point that it is probably wrong to refer to them as such).
In 2013 the state-wide 2PP result was LIB 57.29% (presumably including, by preference flows, the vast majority of the NAT vote) against ALP 42.71%. To bring the closest of these seven seats (North West Central) over the line, the local swing must exceed 11.5%, or a result of LIB 45.79%, ALP 54.21%.
ReachTel's latest polling has the result at LIB 48-ALP 52, which is within the 3 percentage point margin of error of this result. Galaxy has this even worse for the Coalition at LIB 46-ALP 54 which is a rounding error away from colouring North West Central red. And "leaked" internal polling from the Liberal party (which may be skewed to scare LIB voters from straying too far) has the result at LIB 43-ALP 57 which is a swing of over 14 percentage points, within striking range of taking Kalgoorlie (margin vs ALP 16.5%) and Warren-Blackwood (margin vs ALP 15.7%) as well.
An (admittedly very brief) search for polling on the Nationals vs Liberals in WA was not very helpful, but my gut feeling is that the Nationals will widen the gap in these seats. The NAT vote in these seats is too big to be a protest vote that will be shepherded back to prioritising the LIBs by bad poll numbers; it is, instead, a genuine vote for an alternative party. Further the preference shenanigans in WA have been strange this cycle, with the LIBs preferencing One Nation above their Coalition partners and the NATs retaliating by preferencing the Greens, of all parties. I think the NAT voters will feel this betrayal more than the LIB voters, and this may shore up the NAT vote further. Therefore, I think we can call all 6 above NAT seats safe from the LIBs (with the possible exception of Kalgoorlie, which only became NAT last election and has a 3.2% margin against the LIBs, but then again has a stronger history with the ALP than either Coalition party).
Geraldton, however, probably has a strong enough lead to hold off a swing to the NATs, though historically has accepted ALP candidates quite well too.
This means, with the possible exception of a three-way race in Kalgoorlie, we can probably assume no real shift in seats (though perhaps a slight pro-NAT shift in raw numbers) between the Coalition seats.
The other spanner in the psephological workings in WA is, of course, the eternal phoenix that is the One Nation Party. Early polls suggested ONP might reach around 13% of the primary vote, though more recent data suggests this is around 8 or 9%. I am very cautious about all of these numbers as One Nation is the kind of out-there, Trump-style, media-grabbing anomaly that repeatedly threw us and polling companies off in 2017. However, even if ONP performs better than either of these estimates, it's only shot at a lower-house seat is to out-perform a major party and sweep up the preferences. On the latest data the Greens are doing better than ONP, and would naturally benefit from ALP preferences while the preference deal with the LIBs may have tarnished ONP with NAT voters (besides which, out-polling the NATs in the majority of seats would not net a huge pool of preferences to draw from in the first place). The only option, then, seems to be for the ONP to outperform the LIBs. And any seat where the LIBs are coming vaguely close to ONP's primary vote numbers would have to be a safe ALP seat to begin with--probably not good hunting grounds for ONP preferences. Thus, with caution and a healthy respect for the oft-written off right-wing media-loving outsiders of this world, I would suggest ONP is unlikely to win a lower house seat.
With no seats passing to ONP, no real change among Coalition parties and a Greens vote so low in most of the mining-dependent West that I haven't even really addressed it, the real contest seems to remain on the Coalition vs ALP scale. We can return to out initial pendulum, then.
Starting with the most extreme polling, the "leaked" Liberal polling, a swing of 14% against the Coalition parties would look like this:
|North West Central||NAT||11.5%||-14%||2.5%|
This gives the ALP 18 formerly LIB seats and two NAT seats; the 20 seats referred to in the article that is also the source of this polling data. Kingsley also remains in doubt. In reality, I am not sure the NAT seats of North West Central and Pilbara would suffer the full brunt of this swing as this data from WAtoday has the NATs actually increasing their primary vote to 6.4 (from 6.05 in 2013) separate to the LIB decline to 32.5 (from 47.1). Thus the swing may be only minor away from the NATs or even in their favour.
Under the Galaxy and ReachTel polling the NATs would retain both of these seats, and I am happy to call these safe retains for the NATs at this stage.
By comparison to the extreme swing of 14% explored above, the most mild is the ReachTel poll's suggested 9%, which would look like this:
|North West Central||NAT||11.5%||-9%||2.5%|
Let us then take these as the outer limits of possible results; and seat that is predicted identically in the two preceding tables can be called at this point, along with the previously mentioned retention of North West Central and Pilbara for the NATs. I have also given Kingsley to the LIBs as it is a tossup only under the "leaked" data, and Liberal under any more mild swing:
|Electorate||Current Incumbent||“Leaked” Poll||ReachTel Poll||Prediction|
|North West Central||NAT||2.5%||2.5%||NAT|
Already things are looking bad for the Coalition; the ALP has 29 seats and the LIB-NATs have 19 (12 and 7 respectively) with 11 undecided. If the ALP wins any one of those 11, they have a majority. For those curious, the Galaxy poll predicts a swing of 11%, which broadly agrees with the ReachTel results, but puts Wanneroo as a tossup and gives Bicton, Joondalup, Kalamunda and Southern River to the ALP.
Since we don't allow ourselves tossups any more, history is as good a method as any for flipping the coin. Darling Range, has been a Coalition-held seat since being re-established 1977, and indeed through all of its historic incarnations, with one exception of a Liberal-turned-Independent from 1990-1993. Dawesville has been Liberal since it was founded in 1996. Kalamunda has only been around since 2008, with a historical predecessor from 1974 to 1989, but has always been Liberal in those periods. And Murray-Wellington (also called Murray in four of its previous incarnations) has been held by a Coalition party or their predecessors since 1904 with one brief ALP interruption from 1989 to 1993.
|North West Central||NAT||NAT|
Unfortunately the other seats have very mixed histories, except Bicton and Burns Beach which are both new this cycle. (Bicton is formed from old portions of Alfred Cove and Bateman which were historically Liberal, and Fremantle and Willagee which were historically ALP; Burns Beach is formed from Butler--historically ALP--Ocean Reef--LIB--and Joondalup--mixed).
I will think about how best to break the remaining ties, and report back.