Thursday, May 28, 2015

Would Wood Waste Waste The Seat Of Franklin?

There's a law called Betteridge's Law that says any headline that ends in a question mark can be answered with the word "no".  This article is certainly no exception.


Regrowth after logging and burning, showing burnt "wood waste" in the foreground.

It's a polling trope as old as the hills.  Some issue that virtually no-one was hitherto believed to know about or care about is suddenly the subject of startling polling revealing that Party X needs to support it or brace itself for double-digit vote losses.  It's a concept that thrives on the unhealthy symbiosis between activists/lobbyists and journalists (the journalist gets a free story complete with new polling while the activist/lobbyist gets their press release put out as news, usually with no outside scrutiny of the polling involved).  And it seldom if ever amounts to the proverbial hill of beans.

The latest in this unfortunate genre has been an article (Bill Shorten faces a Tasmania Wood Waste Wipeout - Google for article title if paywalled) declaring that polling shows Labor headed for a dire fate in Tasmania if it opposes the inclusion of wood waste in the Renewable Energy Target.  This is based on a ReachTEL poll conducted for the Australian Forest Products Association (AFPA).



"If it opposes what"?

Many readers have probably never even heard of this issue, so I will try to explain.  When forests are logged, a lot of "wood waste" is left behind on the forest floor.  This includes bark peeled off logs (especially in piles at log-landings), sawdust and wood shavings.  It also, however, includes wood from understorey species or that is not usable either for sawlogs or pulp.  This stuff may be either left on the forest floor (for instance during selective logging without burns) or else set on fire during regeneration burns required to renew wet forests that have been clearfelled.  (Contrary to myth, the burns don't generally remove all the wood waste, but they do reduce its volume.)

Interest has now developed in collecting wood waste and burning it for electricity, and the question is whether this should be counted as a form of renewable energy and encouraged accordingly.  Proponents argue that this is making good renewable-energy use of an otherwise useless material, reducing the scale of regeneration burns which can cause annoyance and pollution, and at the same time reducing greenhouse gas emissions from fossil fuel burning.  Opponents argue that it is not much different to fossil fuel burning anyway and would become a covert driver of logging that would otherwise not occur, by making uneconomical coupes economic.

Invertebrate research scientists (and that would be me) say that there's not been nearly enough research on the impact of removing wood from harvested coupes - after all the fact that so many things live in that stuff is one of the reasons why anti-logging arguments are often so overstated - and suggest someone should pay us to do more.  Everyone ignores us, and some acceptable percentage of wood waste to leave behind for little critters is handwaved out of goodness knows where.

The polling - voting intention

The most potentially interesting aspect of this polling is that it includes reports of voting-intention polling for four of the five Tasmanian seats, and they do look less than stellar for the ALP.  The polling, conducted by ReachTEL, was of "about 680 voters" in each seat.  Sid Maher's article for The Australian doesn't give the full results, and they are yet to be released anywhere else.  However, what we have is as follows:

* Franklin: Labor 31.1 (-8.8 since election), Liberal 39.7 (+1) Greens 17.4 (+5.2)
* Bass: Liberal 46.7 (-1.2), Labor 31.1 (-3.6), Green 9.8 (+1.9)
* Braddon: Liberal 46.2 (-0.7), Labor 32.3 (-5.3), Green 6.9 (+1.7)
* Lyons: Liberal 44.4 (=), Labor 31.5 (-5.3), Green 10.9 (+2.6)

The first thing to notice here is that the combined votes for the big three parties in each case are 2.6 to 4.3 points lower than at the federal election.  But at the federal election, the Palmer United Party averaged 7% in these four electorates, and since the election its support in the state (as with everywhere else) has collapsed to a rump of 1-2%.  So the totals for the big three should be higher rather than lower.

That they were not higher suggests either that the percentages are of votes including Undecided, or perhaps that some other voting intention option was given (though it's not impossible there was just a very high Others vote).  Curiously, the breakdown of support in an issues question posted by AFPA in their press release shows voters broken down into categories of Liberal, Labor, Greens and Swinging.  At this stage I do not know if these were the options presented to voters but to be comfortable with using these results at all we really need to see the full breakdown for the voting intention questions.  (Update: I have been advised that the poll included Ind/Others and Undecided as voting intention options, so "Swinging" might be AFPA's interpretation of one or more of those.)

The other important point here is that while ReachTEL polls correctly predicted the winners of every Tasmanian seat in 2013, they did so with errors of 4-5 points in the Coalition's favour on a 2PP basis.  At state level those ReachTELs including an undecided option were significantly more Liberal-friendly than those that didn't. As released, these numbers suggest small to substantial (<4  points) 2PP swings to the Liberals (though Labor would still retain Franklin). But if we adjust them by four points then Labor would hold Franklin easily, and might recover Lyons, while there would also be a swing to Labor in Bass and Braddon. So these results might be reasonably good or distinctly bad for Labor; even with full data it would be hard to know.  (For comparison the current Bludger Track estimate for Tasmania is a 1.1% swing to Labor but no seat gains.)

The increase in the Greens vote in Bass, Braddon and Lyons is plausible - it is consistent with the national trend.  The large increase in the Greens vote in Franklin is also plausible as the party polled over 20% there in 2010.  Its vote collapsed in the seat in 2013, most likely because its candidate was charged with drink-driving.

(I note that in this case the commissioning source's main interest in the poll is the issue questions. This poll is not from a source that would be likely to take many such polls and only release the best ones, as a party, union or major lobby group might do.  So I don't see any reason to doubt the voting intention polling, which was undoubtedly asked first.  However to make real use of it we need to see the full results.)

The polling - wood waste issues

The Maher article and the AFPA press release both report results of a question on whether voters for a party would be "more or less likely to vote for a candidate who supported the reinstatement of wood waste in the RET":


So, more or less across the board it's about 41% for more likely and about 14 % for less likely, with Liberal supporters strongly "more likely", Labor supporters a bit over two to one, Greens supporters not strongly either way and "swinging" voters very variable by electorate.  (If "swinging" was indeed used as a voting intention category this is a strong indicator that the Franklin swinging voters were lefties who would likely vote ALP, further softening the voting intention findings above.)

The Maher article claims on this basis "Labor faces net losses of between 13 and 17 per cent of votes if it opposes the inclusion".  There are so many things wrong with this claim that I really don't know where to start!  Firstly, the 13 to 17 percent figures by electorate are in fact the percentages who said they were less likely to vote for a candidate who supported the reinstatement of wood waste, as can be seen from the table above where they range from 13.2 to 17.0.

Secondly even if Maher (or his source) hadn't misread the table, these percentages would, at worst, only be the percentages of Labor votes that could be moved on the issue, not the percentages of all votes.  We'd be talking about a potential swing of 4-5%, not double digits.

Thirdly polls of this type are no good at projecting actual change in voting intention, because the voter who says they would be "more likely" doesn't say how much more likely, and the voter who says "less likely" doesn't say how much less.  In practice many voters only become slightly more likely to vote a certain way, while others simply lie in response to this kind of polling: voters who are locked in to a particular party and couldn't imagine voting any other way will still say they are more likely to support their party if it does something they like.

These polls are especially unreliable when it comes to obscure issues.  Voters might say that they would take such an issue into account when voting if it is put to them in a poll, but chances are they'll forget about it days after doing the poll.  There is just about no issue that is too silly or trivial for at least 30% of voters to say it could influence their vote.  A better way to do that sort of question is to ask voters how they would vote if the Liberal candidate took stand X and the Labor candidate stand Y, but even then voter responses will tend to exaggerate the impact, because voters aren't good at anticipating how they would actually vote if things were not as they are.

Is this a skew-poll?

Fourthly, because AFPA have not yet released the full details of the poll with the wording of all questions and preambles, we don't know yet whether respondents were "primed".  Priming, in these cases, consists of reading a preamble containing information about an issue, that is typically chosen by the sponsor (not the pollster) so as to portray the issue the way the sponsor of the poll would like it to be portrayed.  I infer that some kind of preamble was read to respondents, because otherwise 90% of respondents would look for the  "huh?" button on their phone when first asked about the issue.

Indeed, a complaint about this poll was seen days before its release.  Former Greens MHA Peg Putt released a press release for Markets For Change complaining about "push polling".

I caution readers strongly against taking Putt's word for it on anything.*

I also observe that if this poll did have a dodgy and biased preamble, then it is amusing in a slightly disappointing way to see the boot on the other foot when the Tasmanian environmental movement is one of Australia's leading producers of dodgy skew-polls with loaded preambles. What comes around goes around, etc.

Furthermore, there is no evidence that this poll was a push-poll as such.  I discussed the difference between a push-poll and what I call a "skew-poll" in Underwhelming Result In Wilderness Poll (a piece dealing with yet another of the dodgy enviro-polls Tasmania is deluged with).  The aim of this poll seems to be not to brainwash the public, but to lobby the Labor Party with evidence that its stance risks loss of votes.  As Labor are such big fans of dodgy message-testing polls themselves, it's more than plausible they'd be silly enough to fall for it.

However, it does appear that the poll questions on wood waste at least had some kind of preamble, and I wonder if the preamble could even be part or all of the bit in italics that appears in the AFPA press release:

Sustainably managed harvesting and processing operations in Australia generate mountains of offcuts and organic waste every year. These offcuts include sawdust, shavings and bark. Additional biomass is left on the forest floor, providing mountains of kindling for mega bushfires or pushed into piles to rot away or be burnt to clear the site for regeneration. With the incentive of renewable energy certificates many regional enterprises would be able to convert coal or gas fired power and heating facilities to use the wood waste, reducing greenhouse gas emissions. The commercial forest operators who are members of the Australian Forest Products Association are 100% environmentally certified by either FSC or PEFC (AFS) and plant or sow more than 60 million trees each year. 

If so, the result would be highly skewed compared to if readers had been presented with arguments in both directions or a neutral summary of the issue.

Amusingly, Markets for Change came up with their own "best way to pose a question", which consisted of a preamble that consisted entirely of arguments against including wood-waste in the RET:

"In 2013 the Australian government stated: “Wood waste from native forests was removed from the RET as an eligible renewable energy source in 2011. This amendment was made to ensure that the RET did not provide an incentive for the burning of native forest wood waste for bio-energy, which could lead to unintended outcomes for biodiversity and the destruction of intact carbon stores.”
Do you agree or disagree that this exclusion should continue?"

This would have unfairly skewed the result in the opposite direction.

Other Issues With The AFPA Poll

The AFPA press release reports 67% support for including wood-waste in the RET.  Again, this is not a useable finding without the full poll wording, or if (as seems highly likely) the full wording includes a skewing preamble.  It then comments "Across the four electorates 62.5% of those in support were Labor voters".  I think it's much more likely that the poll really had 62.5% of Labor voters expressing support, since otherwise support from Liberal voters would have to have been very low.

The press release says "A third of Labor voters told pollsters they would actually change their vote at the next election to a candidate who supported the inclusion of native wood-waste in the RET".  If that claim is based on the table as posted above then it is false, and if it is not then it seems unlikely.  (Even if voters said it, it wouldn't actually be true.)

But Haven't We Seen This Issue Before?

Ah yes!  Only a couple of weeks ago the same newspaper published a poll on the same issue taken by the same company and showing the opposite result!  The article by Graham Lloyd (which is paywalled to death but entitled "Forest Waste in the RET is burning votes") reported on polling commissioned by the Wilderness Society of the southern NSW marginal of Eden-Monaro and the Victorian marginal of Corangamite.

This poll was said to have found 24.2% in Corangamite supporting the plan with 50.1% opposed (41.2-23.4 among Liberal voters), with 45% of Corangamite voters including 21.1% saying they would be less likely to vote for a party that included wood-waste in the RET.  Similar results in Eden-Monaro: 29.6-44.6 support-oppose (Liberal 52.4-17.4) with 43% (including 16.7% of Coalition voters) saying they would be less likely to support a party that took such a stance.

When it comes to such issues, Eden-Monaro and Tasmania aren't so different.  The difference between the polls would be largely if not entirely down to question design.  While AFPA have at least released enough details of their poll to enable me to point out some of the obvious errors in the Australian's reporting of it, the Wilderness Society don't seem to have released anything beyond what is in the article.  I'd bet at very short odds that their polling also contained skewing preambles.

The Perfect Balanced Sample

Bernard: Well, the party have had an opinion poll done, it seems all the voters are in favour of National Service
Sir Humphrey: Well, have another opinion poll done showing the voters are against bringing back National Service
Bernard: Well they can't be for it and against it
Sir Humphrey: Oh of course they can Bernard.  Have you ever been surveyed?

And you can watch the rest here.  I have this video permanently linked in the sidebar because it is very close to a perfectly cynical description of how this sort of issue-polling works, and nothing has really changed in the decades since it was filmed.  That said, instead of bombarding respondents with lead-in questions, most sources that want to commission polls just use a biasing preamble instead.

I'll let AFPA off with a small wirrah, given that it at least it has released some parts of the poll and given that the forest industry in Tasmania has been enormously provoked by an endless stream of terrible green skew-polls and can hardly be blamed if it fights fire with electricity:

Wirrah Award For Fishy Polling 
(image credit)
But for The Australian, allowing itself to be suckered into printing this stuff from both sides (and for the sheer ineptness of its reporting on this one), there can be no alternative ...

Porcupine Fish Award For Ultra-Fishy Poll Reporting
(image credit)



----------------------------------------------------------------------------

(* Putt is an ancient enemy from a turn-of-the-century period of one of my other lives that is known as the Snail Wars.  Putt misled the House by falsely claiming that I had used dead shell numbers as part of a population estimate (for a snail that was wrongly considered endangered) and by falsely claiming that I had done a specific report on a snail and then criticised that report.  Despite multiple requests that she correct the record in the interests of my professional reputation, Putt completely failed to do so and in one case even publicly denied she could possibly have got it wrong.  So when it comes to her comprehension of second-hand information, or her commitment to an accurate public record, Putt gets a permanent F- from me.)

Update (31 May): A second ReachTEL of Braddon conducted by the ACTU has come to light, showing the Liberals on 43.2, Labor on 34.2, Greens on 7.6, PUP 1.5, Others 5.5, undecided 7.7.  The undecideds were more likely to say they were leaning Labor than Liberal (41.2:33.5), on which basis the 2PP might be estimated at 52:48 to the Liberals - either a fairly good or fairly bad result depending on your take on any possible skew.

3 comments:

  1. Hi Kevin,
    Is there much information on the methodology used by ReachTEL? I found it interesting that they seemed to have consistently favoured a conservative viewpoint, and I wondered whether perhaps they might call only land lines, and not mobiles?
    If so, the skew to an older audience might be enough to explain the shift to the Liberals in previous polling. Do they publish any kind of break-down of the age structure within the poll?

    ReplyDelete
  2. ReachTEL is a robopoller (so it makes phone calls using a machine with recorded messages; respondents press keys to select from given options). Their federal polling seems to be more or less neutral, and their primary vote estimates were extremely accurate in both NSW and Qld state elections, so it is clear they don't have an across-the-board conservative skew. In polls taken close to the Tasmanian state election (which did not allow the Undecided option) ReachTEL actually underestimated the Liberal vote. ReachTEL does call mobiles (I am unsure in what proportions compared to landlines).

    ReachTEL don't normally publish the number of respondents in different categories. Other robopolls that sometimes publish this information do show that those who actually answer the phone are indeed skewed strongly to the older age groups - and also at least one robopollster gets a two-to-one ratio of females to males answering the phone.

    Those pollsters whose methods attract an age-biased sample get around it by using scaling, so if your sample has half as many 18-24s as it should have, you just weight all their responses double. This increases the risks of random error and there is also the risk that those within such age groups who actually answer the phone might not be representative.

    ReplyDelete