Thursday, January 18, 2018

New Commissioned Tasmanian Polls

Tasmanian state election season is heating up with regular policy announcements (at least from the government) and rumours that the election could be called this weekend for March 3 (what, not the same day as South Australia again?  Surely too good to be true!)

I expect we will have some public polling before too much longer so we can see if the Liberals have recovered from an utter stinker from EMRS late last year, but in the meantime the shady forces of commissioned polling are out there doing their stuff.  This week Tasmanians were treated to not one but two rounds of robo-bombardment.  A diabolically odd anti-pokies question left many scratching their heads (especially pokie opponents) while reports of the warm fuzzy niceties of the other poll on offer sparked Twitter responses like this:

OK, there was actually only one response like that, but this poll even asked voters if they liked Tasmanian political leaders as human beings!  It also asked if voters thought Jacqui Lambie was good at her job, which came as a surprise to me, because I didn't know she had one anymore.  

MediaReach Liberal Poll

Anyway, the long and cuddly robopoll has seen the partial light of day first, and what this is is a MediaReach internal poll commissioned by the Liberals, with a sample of a whopping 3,000.  Methods details are bereft - I've seen a claim it only canvassed landlines, but I constantly see the same claim about other pollsters who ceased doing so years ago - so there's not much more to say about it yet.  

Now, I'm not sure if ReachTEL weren't available (they're prim and proper about refusing dual commissions if they have a conflict in a market) or if the Liberals just don't trust them anymore, but MediaReach was a novel selection indeed.  This pollster hasn't been seen in Tasmania before and its only previous testable public results have been in the NT, where it was out by about five points 2PP in an electorate poll and a territory election poll.  So what do we know about its accuracy in the Tasmanian or indeed any similar context?  Diddly-squat.  Add to that that it's a commissioned poll that wouldn't have seen the light of day had the Liberals not liked the result, and the only weight I can aggregate it at is zero.  Still, it will be fascinating to see how it scrubs up on election day.

Actually, if I did aggregate this poll it wouldn't make much difference anyway.  Oddly giving results to two decimal places (not that there is anything actually wrong with that) the poll has the following results:

Liberal 41.12
Labor 34.29
Green 12.81
Lambie Network 6.19
leaving 5.59 for others.

Compared with my most recent published state aggregate (from the EMRS piece), the poll has the Liberals up 2.2 points, Labor down 2.7 and the rest more or less exactly where I do.  If these results were accurate, they would most likely point to the familiar story of polling over the term reasserting itself - seats could go 12-10-3 or they could go 13-10-2, and it probably comes down to Lyons.  (Speaking of which, has anyone actually heard from Fraser Brindley?  I've seen no media from him since he was endorsed back in April.)

And that's the perfect story for the Government in terms of the line it wants to play: that only the Liberals can win majority government, but that the election is close enough that every vote matters.  Liberal strategists are reported as claiming they could hold up to 14 seats, but this poll doesn't provide support for that.  Assuming uniform swings, they would need about 49% statewide to retain three in Franklin.  

Not only would the Liberals not have released these numbers if they were much worse, but they probably wouldn't have released them if they were much better either, since nobody would have believed them.  The numbers are therefore totally convenient, but that doesn't mean they are wrong.

(The Mercury's report on this poll, by the way, contains a howler.  The results it claims to have been ReachTEL polls in 2010 and 2014 were actually from the real elections.)

The other thing is that not for the first time, the Liberals are finding that their internals have Premier Hodgman doing better in the beauty contest than the nasty stuff we've been seeing from EMRS.  (As noted in my articles on EMRS, if one treated EMRS as equivalent to Newspoll, one would have sound historic reason to declare Rebecca White Premier already, but luckily for Hodgman the two are not the same.)  The MediaReach apparently has Hodgman up 48-41.4, which by Newspoll standards would denote a struggling Premier rather than a stuffed one, but without any benchmark for MediaReach, who knows?

Bass ReachTEL (updates to follow)

As for the commissioned ReachTEL of Bass, I have seen that poll and, ignoring the pokie troll-poll section, it's fascinating.  Want a lucky number? 58. I will update this article (in the evenings because of work) with comments on such of it as sees the light.

Thursday: The first media report is out for this Australia Institute poll, which finds the Liberals with remarkably good numbers in Bass compared to the general run of recent polling.  After distributing the 4% undecided (the media report just gives the raw numbers), the poll has the Liberals on 49.4% in the seat, Labor on 27.6, the Greens on 10.5, the Jacqui Lambie Network on 10.1 and others on a paltry 2.3.  ReachTEL polls prior to Tasmanian elections have in general underestimated Labor substantially and overestimated the Greens.  The evidence on the Liberals from past polls is mixed (severe overestimates at federal elections but underestimates at the last state election).  Taking all this into account, the most likely reading of the poll is 3 Liberal 2 Labor.  (If the poll is assumed to be exactly accurate, then the Greens' Andrea Dawkins might hang on given that Labor has no obvious number 2 candidate and hence could well suffer significant losses on leakage.  The Lambie Network would be less likely to win because they would be more exposed to leakage issues and would have nowhere much to get preferences from.)

The poll asked voters to report whether Will Hodgman and Rebecca White are the best people to lead their parties.  Hodgman smashed it on this question with a 72.2-14.9 net result (51% strongly agreeing) while White - who EMRS showed as super-popular late last year - did a little less well at 53-36.9.  As well as Hodgman getting a rousing endorsement from his own party here, there was also a pretty strong tendency among Labor voters to agree he was the best choice of Liberal leader.  A cynic might say that Hodgman's strong result here reflects disdain for his party colleagues, but I've seen a lot of reflexive partisan dislike of opposing leaders in my time and I'm not seeing that for Premier Hodgman here.

The Advocate also mentions the issues questions with "fixing the health system" at 37.3%.  "Ensuring a majority government" scored 21.5 and "getting pokies out of pubs and clubs" scored 10.9.  For completion, education on 9.3% and "creating secure jobs" on 21% were the other options.  Probably, education was disadvantaged in comparison to the other options by not being described dynamically - it is usually in the low double figures in such polls.   The well would have been poisoned for the pokies component of this question by a more contentious question about pokies coming before it, but on the other hand the issues question could underestimate interest in pokies issues since a voter who supports keeping pokies in pubs and clubs would probably have picked another option.

I have included the voting intention results for this poll in my aggregate with a weighting of 20% for Bass only, and they significantly firmed up a 3-2 result in the aggregate, though the Greens are by no means out of it yet.

Update: 28 Jan: The Sunday Tasmanian reported more "leaked Liberal polling", which I take to mean internal Liberal polling that the party wilfully provided, ie not leaked at all.  The results reported were:

* 37% think the Liberals will win majority government, 17% think Labor will, 32% expect a hung parliament.  Comment: what the public expect is predictively meaningless except in cases of overwhelming results, but the Liberals would be keen to see that Labor figure stay low.

* Hodgman 50 - White 41 for preferred premier.

This seems to be another MediaReach poll, although that is not stated.


  1. Thanks Kev, was just discussing this.

    Cheers mate.

  2. The Mercury Article also has Richard Herr suggesting the Greens will lose a seat to the Liberals in Bass. Surely the Libs couldn't get four seats in Bass?

    1. On my reading it doesn't directly say Herr suggested that and seems more like muddled reporting. The Liberals can't win four in Bass; if the Greens lose there it will be to Labor or possibly (but unlikely) JLN.