ReachTEL: The PUP Surge Has Landed

ReachTEL: Lib 47.3 ALP 24.6 Green 17.1 PUP 7.5 Other 3.7
Outcome if election was held now based on this poll: Liberal Majority (approx 13-6-4-2 or 13-7-4-1)
New aggregate of all state polls: Liberal 14 Labor 6 Green 4 PUP 1
My current forecast: Liberal 13 Labor 7 Green 4 PUP 1
========================================================================

If the new ReachTEL poll is completely accurate, the Greens could even be the Opposition after the next state election.

Well, that was a rather startling introduction, wasn't it?  The idea of the Greens winning more seats than Labor (or even as many) has been one I have until now regarded as an almost totally crazy beast from outer space. (See my reply to a question from Matt Lyons in comments here where I say there is "virtually no chance" of the Greens gaining seats and give them about a 2% chance of winning more seats than Labor).  But  along comes a poll with a quirky electorate breakdown that, if real, would put the Greens within range of fluking six seats to Labor's five - yes, despite a 4% swing against them and Labor outpolling them by 7%. 



The usual disclaimers apply, of course.  It's just one poll, and the seat breakdowns are c. 600 voters each, and it should be seen against the backdrop of Labor easily outpolling the Greens in Denison for some time.  Even on the poll figures, it probably wouldn't happen (but it might).  Plus it's from a state poll that had the same basic Greens vote as the EMRS released a few days back, and EMRS are known to overestimate the Greens' support at election time.  The suspicion is that ReachTEL do so too, as virtually all pollsters do at federal level.  It shouldn't be taken as evidence that the Greens are actually at all likely to win two seats in the division.  But it should be a reminder that against the mainline expectation of a Liberal majority government with seat losses for both Labor and the Greens, there are some very odd black swans flapping their wings around the edges sometimes.  Not too long ago the Greens were at some risk of being reduced to two seats or perhaps even one.   They're still getting clobbered in swing terms, but the seat picture is much less clear. 

And while my article about a smaller-sample size EMRS just a few days ago suggested Labor's tactic of chucking the Greens out of cabinet was falling flat, with no evidence of a significant increase in voting intention, the question this poll raises is whether it might have even ever so slightly backfired.

The poll: Voting Intention

The current ReachTEL is based on a huge sample of 2,912 voters, making it three times the size of this week's EMRS.  It was taken on Thursday 13 Feb, and although it is released only two days after EMRS, the EMRS survey period was 5-8 Feb.  So the ReachTEL data is 5-8 days fresher, at a time when that sort of thing is starting to matter just a bit.

There appears to be a difference in the way this poll (and the recent Liberal ReachTEL) are conducted compared with last year's string of ReachTEL surveys.  That is that "undecided" is no longer a party option; the poll instead forces a selection at the first voting intention question.  I recently took a ReachTEL commissioned by an unknown source and can confirm that there was not an "undecided" option.  This is discussed further below.

The seat breakdowns are: Denison 594 voters, Bass and Franklin 593, Lyons 575, Braddon 557.

Below I give my usual table of the results, the quota breakdowns and my interpretation of where seats would go if this poll is completely accurate:

Compared to the Liberal internal ReachTEL taken in late January, PUP are up 2.7 points, the Greens up 1.8, the Liberals down 1.5, Labor down 1.2 and Others (including Nationals) down 1.8.  The difference in the case of PUP is what is conventionally considered "statistically significant" (not proving that it's a real increase, but at least providing reason to suspect it might be).   Compared to the November ReachTEL with undecided redistributed (and see below on that) the Liberals are down 5.5, Labor is down 0.2, the Greens are up 1.4 and the generic "others" (including PUP) are up 4.3. 

Oh strange new world ... after a completely routine EMRS just a few days back, this poll is bursting with odd scenarios.  If these numbers exactly were repeated at an election, Bass and Lyons would be the same clearcut 3-1-1s they have been for some time.  In Braddon however, the Liberals would gain a seat at the expense of Labor and there would be an unclear fight between the Greens and PUP for the final seat.  In Franklin, we have the same old question of whether Giddings and O'Byrne would hold their seats, which is closer than it looks because the Liberal ticket would be very susceptible to leakage and there are now Green preferences available for Labor.

Denison is the really fun one because more than a third of the seat is giving the majors the raspberry in this sample, creating something on the cards as a possibility in this electorate for some time: with two Liberal, one Labor and one Green safe, a scramble for the remaining seat with nobody with more than half a quota.  For a long time the fourth-party candidate for this scramble would have been an indie, but in the first ReachTEL taken since Barbara Etter's PUP run was announced, it's her.  On these figures Etter would quite likely win - Labor would be damaged by leakage and Etter would get preferences from everywhere.  Again, this is not consistent with other polling in this seat and the Denison sample especially should be treated very carefully.

So if taken on face value, this poll produces 13-14 Liberal, 5-7 ALP, 4-6 Green and 0-2 PUP.  Probably something like 13-6-5-1 would be the most likely outcome.  But it could be, for instance, 14-5-6-0.  Or 13-6-4-2.  In the context of most of what we've been seeing, these are rather strange possibilities, and at this stage based on one poll we should be extra-cautious about them.

But What About The ReachTEL House Effect?

In previous articles I pointed out that in the federal election, ReachTEL polls displayed a house effect in favour of the Liberals over Labor in every electorate, and it was quite large.  But ReachTEL polls do not display that issue at a national level, and this ReachTEL shows a lower Liberal vote than EMRS.

I'm going to give my hunch here: I reckon something's changed.  In evidence for this theory, I've broken ReachTEL state polls down into those taken before the end of 2013 (which allowed an uncommitted option on party intention) and those taken this year (which apparently did not).  For the former group I've used figures with undecided redistributed.  For each ReachTEL I've found the corresponding Liberal and Labor votes from EMRS at the same time, generally by averaging the two adjacent EMRS polls.

I find that in the case of those ReachTELs that included the undecided vote, their Liberal result was 1.8 points higher than the matched EMRS, and their Labor vote 2.7 votes lower (which is saying something since we already know EMRS results knock Labor about).  For those ReachTELs that do not include the undecided vote, their Liberal result was 1.7 points lower than the matched EMRS, and their Labor votes 2.5 votes higher.  All up that's a difference in the size of the Liberal-Labor gap (relative to that in EMRS) that adds up to 8.7 points - and that's pretty similar to the amount by which ReachTEL overstated the gap in the federal election.  Even if this isn't the reason for this shift, it seems too large to just ignore; there may have been some other methods change.

So while I would normally be whacking four points from the Liberals at this point, adding them to Labor, and arguing that this poll produces a hung parliament (because the Liberals stop winning three in Bass and Franklin) I don't think that I should do that in this case.  It could be that these figures are still undercooking Labor to a degree but if so, it probably isn't that much. 

The Green figures probably are still a bit overstated (as in the federal poll) and if so that would take their Braddon seat and second Denison seat off the table.  Concerning PUP, the suspicion is more that the poll could be understating their support (certainly an issue in the federal election, although they are less of a novelty this time around, and the small-sample EMRS poll did not show so much for them.)  My interpretation of this poll then is that probably Labor and PUP would get the seats "in doubt" and the Liberals and the Greens wouldn't.  But these are quite arguable calls.

New Aggregate

Only just put up a revised aggregate for the EMRS poll and now here we go again!  I think the Liberal ReachTEL can be used in the aggregate (cautiously) now that I have a handle on what might have produced differences between it and previous polls.  The new one should be heavily weighted for sample size and recency, but it's also important to get a cross-pollster view.  So I've decided on the following weightings: this ReachTEL 40%, EMRS 22.5%, Liberal ReachTEL 12.5%, all last year's rope 25%.

And here it is:

Again the aggregate insists on giving the Liberals just enough oomph to win their third seat in Franklin, though it's a result I'm still suspicious of and there isn't that much in it.  When Lara Giddings says she's absolutely confident that she and David O'Byrne will both retain, I really find it hard to doubt her.  On the other hand, with the likelihood that the aggregate underestimates PUP in Denison, Labor's second seat there's not nailed down either.  They do still have some sort of hope in Braddon (because of Hare-Clark being about candidates not just parties).  However, I'm giving that one to Kevin Morgan for now, putting PUP on the aggregate board for the first time (in purple rather than yellow, because yellow doesn't show up well.) 

I also think Etter's chances are worth a serious thought.  Previously I wondered why she was confusing her brand by running with PUP not as an indie (and the sight of a former Integrity Commissioner mouthing Palmer slogans about forestry lockups in ads is one I still find deeply incongruous).  But doing some qualitative research on the leftward end of the Wilkie fanbase (translation: scoffing at the Tasmanian Times greener-than-thou peanut gallery) I'm seeing that Etter still has appeal to that end of the market. Her decision to join PUP has dampened it a little, with some finding it a sad reflection on how hard it is to get elected, but it is far from extinguished.

About The Forecast

As I currently slightly prefer the chances of the status quo result in Franklin despite the polling aggregate suggesting slightly otherwise, I've introduced a forecast table on the sidebar.  This will show what I personally consider the least unlikely result from here, taking into account recent polling but also my own views.  It's a bit like predicting the weather - this far out forecasts still have a fair chance of error (witness the federal election!).  If Marti Zucco's challenge against PUP's registration succeeds on Monday, the predicted seat for PUP will vanish without trace.

The Rest Of The Poll: Preferred Premier

A simple story on the totals; all leaders running slightly ahead of their parties (as is expected when voters for other parties are included), but Giddings less so in proportion to the party vote than the other two.  At electorate level, something I've previously noticed comes to the fore again: in Bass, Giddings trails her party and in Braddon and Lyons she is only barely ahead of it, but in Denison and Franklin she is well clear.  Conversely, McKim is well ahead of his party in Bass and Lyons, and least so in Denison and Franklin.  It's possible from these figures that the PUP/Others voters break much more heavily to Hodgman in Braddon than Denison, or that there are far more Labor/Greens supporters preferring him in the north-west than in Hobart, or both.  (I'm guessing it's mainly the former, and quite pronounced. Could be interesting for preferences.)

Leadership Ratings

Finally!  I've been waiting for real ratings on leadership performance in Tasmania for a very long time.  But a word of caution - ReachTEL leadership ratings have a harsher look to them than those of other pollsters.  I believe that about half of those calling a leader's performance "satisfactory" would express a positive rating in a poll such as Newspoll (which actually asks if voters are satisfied with the leader) while about half would give a neutral or don't-know outcome.

By the methods I outline above, Lara Giddings would have an approval-rating netsat of about -12 (37% approve, 49% disapprove), indicating that on a statewide basis she is mildly unpopular.  On an electorate by electorate basis, I would equate these results to -18 in Bass, -25 in Braddon, -21 in Lyons,  +1 in Franklin and +6 in Denison.  I also think that if Giddings was polling statewide results that were much worse than -12, the Liberals would have revealed it when they dumped nasty netsat findings about various leaders from their internals.

On to young William then:

Again, these results may look a bit drab, with scarcely more voters liking Will Hodgman than disliking him.  But given my comments above about "Satisfactory" I reckon he's doing quite well here.  I translate this to a statewide Newspoll-style netsat of +18.5 (52 approve, 33.5 disapprove) with an electorate breakdown of +23 in Bass, +20 in Lyons, a stonking +38 in Braddon, a slightly mediocre +13 in Franklin and a mere +1 in Denison, the only division where Giddings' ratings are better than Hodgman's.  (I think if the Preferred Premier question had been Giddings and Hodgman only, she would have narrowly beaten him in Denison on that one too.)

Issues

Nothing surprising here.  The jobs/economy nexus (usually as separate questions) has dominated previous issues polling.  "Crime and safety" isn't getting much of a run at the bottom, but that won't stop the Liberals flogging that issue.

Labor's Divorce From The Greens

You can see why this is such a "meh!" in terms of voting intention.  Bearing in mind that most people who say issues will move their vote either aren't telling the truth or else aren't much more/less likely to vote a certain way, it wouldn't make much difference even if it wasn't so evenly split.  As it turns out this only gets the thumbs-up in Braddon, and a slight thumbs-down everywhere else.  It's hard to unravel how many of the "less likely" votes are code for hating everything Labor does and how many of the "more likely" votes are code for hating the Greens more.  The next one is much more interesting ...

The Pulp Mill

This one's really funny!  So funny it deserves a separate article.

UPDATE (20 Feb): Very basic details from another Liberal ReachTEL have been released:

ALP 24.6 LIB 47.3 GRN 15.8 NAT 1.9 PUP 6.3 OTHER 3.7

It's a party poll but I don't have much trouble believing it.  The question is how many unreleased party polls there are out there that may show very different results.

What's interesting here - though no difference is statistically significant - is that Palmer United is doing worse in the poll that includes the Nationals, compared to the one that doesn't.  All the others have very similar votes except for the Greens who are down slightly.  There might be nothing in it but the question is whether a certain kind of anti-three-party vote would be competed for between the Tasmanian Nationals and PUP.

UPDATE (24 Feb): Some readers would have seen this in the Sunday Tasmanian.  It's an example of how the media sometimes report the analysis of possibilities badly.  The commissioned report in question does exist (it's not the same as this article as some believed).  The report does find that the Greens could gain seats if they could get their vote to certain levels in Denison and Franklin, and would be likely to do so if they could reach certain higher levels.  But I was not commissioned to report on the chances of any of those ifs coming true.  Indeed, fearing such a misunderstanding I explicitly noted that two Greens in Franklin would be completely inconsistent with all recent polling.

The SunTas piece's use of the term "predicted" is especially unfortunate, since one doesn't predict an event premised on an wildly unlikely conditional. 

To say that my work shows the Greens could win two in Franklin makes no more sense than if the Socialist Alliance asked me what vote they would need to win all 25 seats statewide, and I said 88%, and someone on that basis then wrote that I had showed the SA were on track to dominate the election.

Once more with feeling then: a two-Green result in Denison remains highly unlikely but is a scenario that should not be dismissed entirely. A two-Greens scenario in Franklin, however, is not even remotely realistic.