Thursday, August 24, 2017

ReachTEL: Bob Brown Foundation Peddles A Poll Porky

ReachTEL (commissioned) Liberal 41.2 Labor 33.3 Green 13.1 PHON 4.4 Ind/Other 8.0
Interpretation Liberal 41.2 Labor 37.1 Green 10.4 Other 11.3
Most likely seat result based on this poll 12-10-2 plus one seat undecided between the three main parties
Disclaimer: Polls are snapshots not predictions.

I may get taken off their media mailing list for saying it, but the Bob Brown Foundation have released a grossly deceptive claim concerning their current ReachTEL poll to the Tasmanian press.  Question two of their poll and its results are as follows:

The question design is a bit odd, but I think it is OK.  The results are 38.3% support, 44.7% oppose and 17.1% don't know/not sure.

On this basis the BBF's press release has claimed:

"A ReachTEL poll commissioned by the Bob Brown Foundation shows that a clear majority of Tasmanians oppose Tassal's invasion [sic] of Okehampton Bay with Atlantic salmon fish farms.


The telephone poll, of 1,617 residents across Tasmania last Thursday, revealed that 44.7% of people opposed Tassal's move, 38.3% supported it and 17.1% were undecided. If the undecided vote is allocated proportionately, opposition to fish farms on the east coast is 54% to 46%."

But from the actual polling figures, the first claim is clearly false. "A clear majority of Tasmanians" means more than 50% of Tasmanians.  A Tasmanian is still a Tasmanian even if they Don't Know or are Not Sure about a proposal on fish farms.  They don't softly and suddenly vanish away just because they pressed button 3 on a ReachTEL poll! It might be that if they did form an opinion that they would break the same way as the support/oppose breakdown (53.9% oppose), or they might break completely differently, but at this stage they don't have a view.  In fact, since not having a view on something is clearly not the same as opposing it, the poll  more correctly shows that for now, a "clear majority" of Tasmanians (55.3%) do not oppose Tassal's proposed expansion.

Moreover, even if we did ignore all the don't-knows, there would be enough questions about whether this poll had actually shown a "clear majority" in opposition.  The notional margin of error on the support/oppose responses is 2.7%, but ReachTEL uses a large degree of response scaling that means the real MoE is larger.  Election results have also suggested that Tasmanian ReachTELs tend to skew slightly to the Greens, who are of course the party most likely to oppose fish farm expansion.  So it's far from nailed down by this poll that more voters oppose fish farms in Okehampton Bay than support them.

I'll make it clear here that I don't have any stake in the Okehampton fish-farming matter - to the extent that I care about it at all, I am much more concerned about the risks to the endemic Maugean skate from fish farms in Macquarie Harbour.  As concerns Okehampton Bay, I would have pressed the don't know button as well.  But I am, as always, tired of environmental groups releasing nonsensically puffed-up spin about their polling results to the press.  And since this is a poll about fish farming, here's one you won't be seeing in the Tassal pens  ...

Porcupine Fish Award for Ultra-Fishy Poll Spruiking
Awarded to Bob Brown Foundation (image credit)

Voting Intention

ReachTEL have an excellent record in not allowing commissioning sources to influence their voting intention polling.

After redistributing the undecided voters, the poll's voting intentions are Liberal 41.2% Labor 33.3% Green 13.1% One Nation 4.4% Ind/Other 8%.  (Sample size 1617).  The inclusion of One Nation is surprising since the party appear not to be running in the state (although one cannot be sure), whereas Jacqui Lambie Network are making noises about running (although these may not amount to all that much).  

If this poll is taken on face value, the most likely result is a hung parliament with 12 Liberal 10 Labor and 3 Green, although 13-10-2 is possible if the Liberals split their vote in Lyons evenly enough, but on the other hand it might be the Liberals rather than the Greens who would drop a seat to Labor in Bass.

However there is good reason not to take Tasmanian polling at face value - in past polling from the 2013 and 2016 federal elections and 2014 state election, ReachTEL consistently overestimated the Greens and underestimated Labor.  (On this basis, 13.1% is a concerningly poor result for the Greens.) When this is adjusted for, based on the evidence of those elections, I have the current poll looking something like this:

This adjusted version has the final seat in Lyons as a three-way fight, something which has been showing up a few times now since the leadership change improved Labor's polling.  Rebecca White would be elected first with a large surplus and the Liberals would hope to split their vote as evenly between them as possible.  There would be a race between the top two remaining Labor candidates and the Greens' Fraser Brindley.  The Labor candidates would be likely to receive preferences from minor parties in the event that those parties even contested, or votes directly if they did not.  This would compensate for leakage from White's surplus.  If the Labor candidates managed to beat Brindley then Greens preferences would help them both pick off whichever Liberal was running last, and Labor would win three seats.

On the other hand if Labor's spread between its remaining candidates wasn't close enough, or it lost too many votes to leakage (these are big risks with no obvious #2 and #3 after White) then it would come down to the familiar question of whether the split between the three incumbent Liberals was even enough for them to beat Brindley.  

I have added this poll to my state polling aggregate at a weighting of 12.5% (a discount has been applied because it is commissioned.)  Here's the aggregate:

It is a similar story to the above, except with the Greens as the stronger threat to the three Liberals in Lyons than Labor are.

Of course this still ignores the possibility of high-profile independent candidates emerging - I can't do anything about those until they announce they are running and start to make their presence felt in polling.


I was struggling to believe my eyes when I saw the initial Sportsbet odds for this election! They have Labor at 1.20 to provide the Premier compared to the Liberals at 4.00 and the Greens at 15.00.  There's a case that Labor should be favourites, but I think it is closer than that.  But what was more interesting was the type of government formed odds - Labor minority 1.22 Liberal minority 4.00 Labor majority 17.00 Liberal majority 15.00.  Another hung parliament does seem pretty likely at the moment but I suggest both majorities here (Liberal especially) are too long.  Even if voting intentions do not shift, Hare-Clark is quirky, and the weakness of the Green vote and volatility of the Tasmanian voter base gives both majors a shot at majority government.  (In Labor's case, they need their polling to further improve, as they are not currently in the hunt for more than 11 seats.)

In 2006 minority government was assumed to be inevitable because EMRS polls did not show any party near to winning a majority.  In fact, EMRS had a large undecided rate which proved to consist mostly of soft Labor voters, who snowballed back to the party once it became clear that only it could win a majority.  Bookies had Labor as long as $9 to retain their majority and yet that is what happened - not only that but Labor did it in style and almost picked up a seat in the process.  Whatever the odds of a hung parliament in 2018, it is simply not, as the current odds imply, an almost 90% chance!

Other Questions?

A commissioned poll including salmon questions about fish-farming at Okehampton Bay and also north-western Tasmania and also five questions about the Tarkine (an obvious matter of BBF interest) was reported in the field by multiple sources on August 16.  The current poll was dated August 17 according to the press release.  It is unclear at this stage whether the other poll was the same poll as this one or different.

(NB Five words from The Hunting Of The Snark were harmed in the making of this article.)

1 comment:

  1. $15 for a Liberal majority is clearly way too long. That is pricing in a 5-6% probability of the Liberals winning. Based on current polling very roughly you would expect something in the range of 45% chance of minority government, 45% chance of Liberal majority and 10% chance of of Labor majority (10% is probably too high but if you ran any statistical models you'd be hard press to get less than 10% due to the uncertainty and long tail in this probabilistic distributions).

    Ultimately Liberals are way too long and hung Parliament is too short. Similarly the likelihood of having Rebecca White as Premier is too short and Will Hodgeman as too long. I'd probably have Will Hodgeman at slight favourite (even if I would rather it go the other way).