Sunday, September 14, 2014

ReachTEL: Liberals With Solid Lead

ReachTEL (Tas State) Lib 48.1 ALP 28.7 Green 16 PUP 3.7 Other 3.4
Interpretation Lib 49.2 ALP 32.2 Green 13 PUP 2.7 Other 3
Outcome if election held now based on this poll: Liberal Majority (approx 14-8-3)
Current state "nowcast" aggregate of all polling: Liberal 14 Labor 9 Green 2

Since the Hodgman government's emphatic majority victory six months ago, two Tasmanian state polls have been seen.  The first EMRS (No Honeymoon For Liberals) showed very little difference to the election result, while the second (Closest Gap In Four Years) was underwhelming for the Liberals, though it still showed them in majority-winning territory.

Now a new ReachTEL robopoll published in today's Sunday Tasmanian, with a sample size larger than the two EMRS polls combined, provides some further evidence that Labor support is rebuilding, while suggesting the second EMRS could have been an outlier.  The poll is rosier for the Government than the second EMRS - on basic voting intentions - and suggests that much of the ALP rebuild is coming at the expense of Palmer United.  The poll also provides fairly good results for the Government's first Budget, but an indifferent response on job creation.



Voting intention

First, the basic primary figures:


I'm about to explain why we shouldn't take these at exactly face value, but if we did, the result of an election "held right now" would be a majority win for the Liberals with 14 seats to 7-8 for Labor and 3-4 for the Greens.  The Liberals would lose their fourth Braddon seat to Labor, the Greens would gain a seat from Labor in Lyons, and the Greens would be at risk of losing their new leader's seat in Bass to Labor.

We should bear in mind not only that each of the individual seat results has its own margin of error (in most cases about 3-4%) but also that with 25 individual seat results being measured at once, it's much more likely than not that at least one of them is out by more than that MOE.  Of these the two that leap out as likely to be victims of random sample error are the Labor vote in Lyons and the Greens vote in the same electorate.  It makes no sense that the Labor vote would have gone down still further in Lyons while going up in the state in general.  It also makes little sense that the Greens vote in Lyons would be above the state average having been 2.4 points below it at the election, and now with the party lacking an MHA in the electorate.  So I would not take the apparent recapture of a Green seat in Lyons seriously yet.

What did the state election tell us about ReachTEL?

The state election provided an important opportunity to get some information concerning the "house effect" of ReachTEL polling at state level.  A ReachTEL poll was released on the weekend prior to the election (ReachTEL: The Campaign Has Changed Nothing).  On election day the Liberal vote was 3.8 points higher, the Labor vote 3.7 points higher, the Green vote 4.4 points lower, the PUP vote 1.7 points lower and the vote for others 1.4 points lower, than that final ReachTEL poll.

I believe that there was some vote movement from the Greens to Labor in the final week, based on a feeling that the election was a done deal and concern that Tasmania might be left with a hopelessly split or uncompetitive Opposition.  However, if it was as much as four points then it must have come extremely late, since Newspoll also underestimated Labor and overestimated the Greens (as did EMRS, but that's business as normal for them.)

What makes this tricky is that in the federal election (but with different question design via the inclusion of an Undecided option) Tasmanian ReachTELs overestimated the Liberal vote, even though it probably increased between when they were taken (which in most cases was well out) and the election.  And in general federal polling, ReachTEL showed relatively little divergence from other polls.

What we do know is that in all these different cases, the Green vote was overestimated, as it is by almost every other pollster, especially of Tasmanian state politics.

Trying to estimate how Tasmanian state ReachTELs relate to underlying electoral reality is a complex and inexact task, but using their state (weighted most heavily), Tasmanian federal, and overall federal polling (plus a degree of stability weighting for the possibility that observed shifts largely reflected campaign events) I've come up with the following working adjustments: Liberal +1.1, Labor +2.6, Green -2.5, PUP -0.9. 

Applying these to the current poll the result looks like this:


As noted above, the third Green seat shouldn't be taken for granted.

Voting intention aggregate

In the EMRS articles I presented a state voting intention aggregate based on those two polls and the state election result.  It's time to throw my interpretation of this ReachTEL into the mix.  I don't want either EMRS or ReachTEL data to over-dominate the aggregate, but the ReachTEL data are both fresher and based on a larger sample size (albeit sampled on a single day.)  I've therefore weighted the new poll at 60%.  Here's the new aggregate:


In this version of the aggregate the Lyons sample puts the Greens in what looks like a winning position in Lyons, but based on preference flows at the last election it is actually more likely Labor would get the seat.  In Bass the Greens would have the same problem as discussed in the last EMRS piece: that unless their primary vote lifts to much closer to a quota, they can lose their seat simply because of changes in the Labor vote.  In Franklin the Liberals would be at great danger of dropping a seat because of leakage, but with three incumbents to Labor's one (rather than 2-2 at the last election) it might be as much of a problem for Labor.

13-10-2 would be a possible outcome on these figures, but a convention I'll be adopting is that when I call a seat seriously unclear, I'm assuming it retains its existing pattern as far as possible.  Thus my seat "nowcast" (a statement of where the cards would fall based on current voting intention, and not a prediction) at this very early stage is 14 Liberal, 9 Labor, 2 Green.  It means nix predictively so far from the next election, but the danger for the Greens in Bass is real.  They shouldn't hide behind rosy scenarios and ignore the evidence that polling in this state almost always shows them doing better than they are.  If they cannot lift their primary vote substantially from the 2014 result, they will likely lose another seat.

The PUP Bubble

This poll provides further evidence (if any was needed after the derisory 1% in the EMRS survey) that PUP are tracking below their state election result in Tasmania.  I have read any number of interstate commentaries claiming that Jacqui Lambie is the new Brian Harradine and implying she is set to be a major player.  It doesn't seem anyone writing this stuff is bothering to talk to Tasmanians about their views of the Senator or seeing any need to have a look at PUP's polling in the state.  While there are differences between state and federal voting, if Lambimania was really sweeping the island one would expect the PUP vote at state level to be booming, yet it is doing anything but.  This could change down the track if Lambie can produce more real achievements for the state and temper her proneness to media accidents and shrill Hansonesque outbursts, but let's wait til that happens before calling it. 

Preferred Premier

Here are the ReachTEL Preferred Premier scores:



And by party:


These are not hugely different to the two EMRS results thus far (54-22 and 51-25, Hodgman-Green in each case) but there are two methods differences to consider.  The EMRS question allows respondents to be undecided, but the ReachTEL question allows a Booth option and disallows an undecided response.  Comparing the Greens voters responses between the two, it turns out that they're more likely to pick Kim Booth if he's an explicit choice, than when they're just at first asked to pick between Hodgman and Green.   It also looks like there are voters for both majors who will pick their man if forced to make a pick, but otherwise refuse to pick one.  The constant theme across both polls is that about a fifth of ALP voters prefer Will Hodgman as Premier to Bryan Green.  This is still not unusual for a new Opposition Leader against a new Premier who was very well known prior to coming to the job.  However, I normally find that the advantage to incumbents on these questions is lower in robo-polling than phone polling, so Labor would want to see that figure coming down over time.  A strong result for Hodgman then, but not all that surprising.

The Budget


If you add up the Very Goods and the Goods on one side, and the Poors and Very Poors on the other, you get a net -1.4, but ReachTEL polling doesn't really work like that.  The use of "Satisfactory" as a middle option tends to attract a 50-50 mix of mildly positive and neutral/indifferent sentiment, and this can be seen by comparing ReachTEL approval ratings with those of other pollsters.  To get a net rating for a ReachTEL poll I count half the Satisfactory results as positives, and on that basis I get a statewide +13.2 rating for the Budget. (By electorate Bass +14.8, Braddon +9.1, Denison +3.5, Franklin +17.9, Lyons +20.7).  So I think this poll shows a fairly positive response to the budget.

By party there's a certain level of indifference from each, but otherwise the response is very partisan:


Job Creation


The job creation question however, even allowing for my comments above about the "Satisfactory" option, isn't so pretty for the new government.  The converted net rating comes out at -4, though this mostly results from more than half the Denison respondents giving a Poor or Very Poor rating.  While it's tempting to conclude that this is because public servants in Denison are angry about job cuts, I think the main explanation is partisanship.  Denison voters are more likely to vote Labor or Green and Labor and Green voters statewide have very negative views of the government's job creation performance.  It turns out that alone just about explains the Denison figures:

Most likely very few public servants concerned about losing their jobs would be Liberal voters anyway.

On this question Liberal voters are more likely to pick "satisfactory" than Labor and Green voters, who tend to have strongly negative views of the government's perfomance already.  The electorate that is giving the government a bad rap here even taking the party breakdown into account is Lyons, where employment problems resulting from the near-total local collapse of the timber industry remain severe.

I've not copied most of the demographic breakdown stats because they largely tell us what we already know - older voters are conservative, male voters are more likely to vote Liberal or PUP and female voters Labor, Green or other, and under-35s are a little more pro-Liberal than 35s-50s. 

Other questions

The poll also includes a question on extending state schools to Years 11 and 12, which has strong support from Liberal voters and weaker support from the other parties, with all electorates supportive (low-retention Braddon least so, partisan breakdowns notwithstanding.)  Matt Smith (Sunday Tasmanian) has detailed discussion of this question with responses from a range of commenters on education and employment matters.

Finally, a question on development in national parks and a question on the Mount Wellington cable car have results likely to be published later this week.  I have seen the results but they are embargoed until published.  The cable car poll is of great interest to me because while public opinion on it has been "surveyed" in various ways, none of them have ever been credible polls, and most of them have not been polls at all.  100+ years after the idea was first mooted, we finally have a real poll on it!
A full article on the cable car poll should appear here in the upcoming week.

4 comments:

  1. Looks like Kim Booth is relatively popular, polling ahead of the party, except in Denison. Is that common for preferred premier results including the Green leader?

    Very interesting to see the strange Green Bass vote, especially considering all the special attention they've put into Launceston: Based their second Senator there, State Leader there, rented a campaign HQ and ran one of their better doorknocking campaigns, and they're having their state conference there next weekend. Isn't the Green Bass vote just as likely to be wrong in this poll as the Lyons vote? Didn't just get the columns mixed up, did they?

    Anyway, glad to see another Tasmania poll regardless. EMRS is nice but too few and far between. Will be keeping a keen eye on that Green vote.

    ReplyDelete
  2. All the ReachTELs I can find for Nick McKim had him running ahead of the party, by varying amounts but similar on average to the gap in this case. It's to be expected because those who pick someone other than the three main parties have to prefer someone for Premier.

    The Green vote in Bass at this poll is actually more or less the last election result. Accounting for house effect, the poll's pointing to something slightly less but still within MOE of the election result, and it's the only electorate sample for Bass we have so far. Many of the things listed were already factored into the election result and others don't really affect voter intention. We'll need to see a few more of these to get a better feeling for how evenly the Greens vote is moving by electorate.

    ReplyDelete
  3. How is the state government comparing to poll results for the Groom government in 1992. The election result back then was similar to what we got in March this year and it was produced from similar circumstances although I was too young to remember actions that immedietley followed. I was just wondering if the electorate back then maintained high support for that Liberal majority government months after that election ? Thanks. I can't wait for the cable car poll, should be interesting.

    ReplyDelete
  4. Alas I don't have any polling records for the Groom government in 1992-3. I suspect Morgan were polling Tasmania during that time, doubt that anyone else was.

    ReplyDelete