A New Species Of Strangeness In The West

It's a familiar polling story.  The long-serving government is looking rough around the edges and the Opposition has a moderate but handy polling lead.  The Premier is on the nose and has been rating badly for years, the Opposition Leader is well regarded and leads as preferred Premier.  There are rumours and rustlings, and finally a challenger indicates potential interest in the job, should his/her colleagues so desire it, and says that the current leader can't win the election.  We all know how this story ends.

Only in Western Australia, it's the Opposition Leader, not the Premier, who has been informally challenged - and even more strangely, the putative challenger isn't even in parliament.  In an extension of the Campbell Newman doctrine (in which a new leader can be drafted by the party from outside the parliament in order to take over), it is now possible for someone to raise an interest in the leadership when they:

* aren't in the parliament or serving in any current political role
* have no seat for which they are reasonably assured of preselection
* aren't being drafted by the party's current organisation or the parliamentary party, and whether anyone involved in the drafting is even a member of either isn't totally clear



The putative challenger, Stephen Smith, is, of course, a former Federal Minister.  Smith was well enough regarded as Foreign Minister and then Defense Minister in the Rudd-Gillard governments, but the former is a role in which even duds often shine.  At some points Smith was even talked about as a possible "nightwatchman" Prime Minister to resolve the Rudd/Gillard conflict.  At the moment, it looks like Smith's challenge is being emphatically shut down, but should Opposition polling worsen, he has his foot in the door for another go down the track.

Quality public polling in Western Australia has been scarce since I covered it in an article called Unpopular State Premiers Have Dire Historic Fates.  (That article pointed out that of sixteen Australian state premiers to have polled Newspoll netsats of -20 or worse, only one has led their party to victory at the following state election.  With the subsequent defeats of Campbell Newman (worst netsat -24) and Lara Giddings (-33) the sixteen has now become eighteen.)

The following chart shows the pattern of recent WA state Newspolls:

The pattern is relentless: for over two years Barnett has been consistently unpopular, McGowan has been consistently quite popular, and McGowan has been consistently considered "better Premier" (which is amazing given how much this indicator skews to incumbents and in the context of a close 2PP).  More concerningly for the Barnett government, they have fallen behind in the recent 2PP Newspolls (as hinted at in the previous piece: bad netsats can be a leading indicator of worsening 2PPs).  I treat these with slight caution because of the changes in the Newspoll brand, which at federal level appear to have produced a persistent lean to Labor worth about 0.6 two-party preferred points.

There is another pollster, of a sort.  Morgan's ten state SMS polls taken through 2015 (with samples usually around 400-600) showed an average 2PP of 50-50, with McGowan leading as preferred Premier by about 59-41 (Morgan has no "undecided" response).  This was followed by an apparently rogue February 2016 sample with the Liberals supposedly leading 54.5:45.5 though with McGowan still preferred Premier by 56.5% to 43.5%.  I have very little trust in the Morgan SMS method, with Morgan producing the least accurate final polls for both the Victorian and NSW state elections (though ranking midfield on primary votes for Queensland), and with Morgan's campaign-period polling for the Victorian state election extremely bouncy.  I fear that Morgan is especially prone to sample-pool based distortions in the less populated states.

What is notable in both Morgan and Newspoll series is that there was very little if any positive change for the Coalition upon the replacement of Tony Abbott with Malcolm Turnbull as Prime Minister.

The case for Smith as an alternative to McGowan is based on McGowan supposedly lacking flair and on doubts that the party can achieve the large swing required to win the election.  There is no evidence that anything much about Opposition Leaders matters so long as their party is leading the 2PP and they are not dangerously flaky.  Cue the usual mantra: governments lose elections rather than oppositions winning them.  Of course, oppositions can lose elections off their own bat as well, but having a leader who's seen as boring isn't enough to do that by itself.

On the latter point there has been a lot of talk about the 10% swing required for victory, but it is worth looking at this a little more closely.  Following a redistribution (see the Poll Bludger 2PP margin estimates here) Labor gains the new notional safe seat of Baldervis, but two of its current seats (West Swan and Collie-Preston) become notionally Liberal.  A uniform swing of 10.1% claims the tenth seat to fall, Joondalup.

But let's look at what the pendulum would look like post-election if Labor achieved that uniform swing in every seat:

I've just shown here those seats that would be closer than 10% in either direction.  The Coalition would have won four seats to Labor's one on margins of below 1%, and eleven to two on margins of below 3%.  In practice, because swings vary from seat to seat, this kind of thing is highly improbable.  A 10.1% swing would be likely to actually take a few of the seats that the uniform swing projection would give to the Coalition.

This isn't one-way traffic - the Coalition has a sophomore effect on its side in the seat of Balcatta, which makes that seat closer than it appears on paper.  However, taking into account sophomore effects, a quick probability-based model I've done suggests that the swing Labor needs for a 50% chance of winning the 2PP in the majority of seats is not 10.1% but only 8.3%.  (I've assumed a standard variation in seat swings of 3.5% and that double sophomore effect in seats where the Coalition incumbent won from the Labor incumbent in 2013 is worth just over 2%.  If a smaller variation is assumed, the swing required increases slightly.)  (See also William Bowe here who argues on geographic grounds that even 7% could be enough).

This makes sense because the Coalition is not all that protected by sophomore effects (most of the seven seats won from Labor in 2013 would fall easily on a substantial swing back, while one, Pilbara, is reasonably safe).  So unless there is a serious geographic skew in the Coalition's favour (like that to Labor in South Australia) it should be possible for Labor to win with not much more than 50% of the two-party vote.  My modelling gives Labor a 2PP win target of about 51% 2PP, as opposed to the crude uniform-swing models which imply a 52.8% target.  This is yet another example of how simplistic use of the Mackerras pendulum is a problem - in this case one that may make people think it is much harder for Labor to win than it probably is.

The other point here is that the Queensland and New South Wales elections both showed that large swings back to Oppositions happen.  In both these cases, Labor was in federal government (and very unpopular) at the time of the lopsided state result.  Exactly the same applies here.  In these days of declining long-term party loyalty, there is hardly any correlation between 2PP results from one election to the next even at federal level.  At state level, with federal factors such a massive contaminant, the idea that the 2PP dice will have any memory from an election held four years ago is simply ludicrous.

We don't have the polling to say directly that the Barnett Government is doomed.  The Newspolls have Labor only marginally above the win target over the last three polls, and such a deficit could well be recovered or even a product of polling error (with only one well-regarded pollster serving the state it is hard to say).  The personal ratings history suggest that Barnett will himself either lose or be replaced by his own party, but all these rules can have exceptions.  If Labor is in office federally then the current polling will probably very soon become irrelevant.

But all things considered, Labor's position is one most state Oppositions can only dream of.  If this is really time to dump the leader, then when is it not time to do so?  How much should an Opposition Leader be leading by for him not to be replaced by someone who has not even been in the Parliament while he has been at work?  This seemed not just like the coup culture that saw Kevin Rudd removed on flimsy psephological evidence, but a feral outgrowth of it pushed to new extremes because it can be.  The rolling of leaders who are failing is one thing, but the idea that leaders should be removed because it is not certain they will win and because someone else might (no public polling evidence for this even existing) is bizarre.

Update: Polling (Friday 18th March): A new 7 News ReachTEL has been reported showing McGowan preferred as ALP leader by 48% of voters compared to 28% for Smith and 24% for Alannah MacTiernan (who also isn't in state parliament right now, but has a long history of popularity).  It would be useful to see a party breakdown here to see how many of the votes for the alternative choices come from mischief-making Coalition supporters.  Hopefully one will arrive in the next few days. A preferred Premier poll between Barnett and McGowan is coming tomorrow and I expect McGowan to have a substantial and perhaps even large lead. The voting intentions results will be interesting too.

Update midnightish: Yes McGowan is up 61-39; many results are here.  The respondent-allocated 56:44 to Labor on primaries is rather generous; provisionally I get something more like 53:47 by last-election preferences, but still handy confirmation of the Newspoll results, albeit within a new state market for robopolling.  (William Bowe who is obviously far more familiar with WA preference flows than me since he lives there gives "a bit under 54%").