Wednesday, October 1, 2014

Recent Polling In Four States

NOTE re Vic polling (28 Oct): Three new Vic polls have appeared in recent days: a 52-48 to Labor from Galaxy, 53-47 from ReachTEL, 52.5-47.5 (with a much too high Green vote) from Morgan SMS.  However Fairfax Ipsos and apparently Newspoll are in the field so we may as well collect the full set.  I am busy with Hobart Council for a few days but will have detailed coverage of Victoria shortly.
======================================================================

This could be a dull week in federal polling, but fortunately new state polling has been released in four states.  The last article on polling from states other than Tasmania was Recent Victorian and Queensland Polling, over a month ago.  This one will cover what has come out since, and updates will be added for anything else that surfaces in the next few weeks.  Beyond that we will be getting seriously close to the Victorian election and polls should become more common.

Immediately after this article was written a Morgan poll including data from all states was released.  Comments about it appear at the bottom.



Victoria: Snack-Sized Polling Morsel Added
At this stage the only new public material in Victoria is an Essential poll that has been reported by The Age as having Labor ahead 52:48.  This compares to readings of 54:46 from Nielsen in July, 52:48 from Galaxy in mid-August and 55:45 in the two-month July-August Newspoll.  I believe the new result is from a four-week sample period.  

The Age compares the 52% 2PP to a 53% result attributed to "March".  This caused me some confusion as I was aware of a very different March result from this pollster, but I've been advised that the "March" result the Age refers to was in fact a rolling sample over the second half of February and the first half of March.  To illustrate how arbitrary it is as a comparison point, monthly Victorian samples from Essential between November 2013 and May 2014 ran (figures are for the Labor 2PP) 48, 52, 50, 55, 50, 51, 53.  (Those between then and now have not yet been released.) What we can say in general is that Essential hasn't bought the size of leads seen in some of the other polls.  The question then is what significance we should attach to it not doing so this time either, and indeed returning a result right on its average for the first five months of the year. I would say, not very much.

The Essential poll is reported as showing very poor ratings for the government on job creation (15% rating it good to 41% poor).  Admittedly this may be a hard question to rate well on, with the new Tasmanian government recently polling quite mediocre results on this issue after only six months in office.

The Australian did one of their seats-in-play pieces in which they talk to insiders about party polling, insiders make unreliable noises and the authors then scrutinise various tealeaves.  Morwell (a notionally safe seat where there is a backlash over a coal mine fire), Ringwood, South Barwon, Mordialloc, Bentleigh and Carrum were classified as at risk for the Coalition (though the evidence for Morwell came with reservations), and Ripon (now notionally Liberal), Bunniyong, Eltham and Yan Yean for Labor.  Frankston, won by Liberal-turned-independent Geoff Shaw, is at obvious risk of falling to Labor.   Both parties are said to agree that Labor is ahead but that the party's lead has fallen, amid softening anger against the Abbott Liberal government.    I note this seat list so we can see how it scrubbed up after the election.

Headline betting rates are currently implying winning chances for the Opposition of 67% (Sportsbet) and 75% (Sportingbet).  I think that the latter at least is too high.  (Peter Brent in Careful With The Vic Pendulum goes further and rates it closer to even than 75%.) Sportsbet also has odds on every seat, but in a rather user-unfriendly form that makes whizzing through them all quickly difficult (at least on my browsers).  Published polling still isn't telling us much about the shape of the likely outcome other than that the government (which if historic patterns persist will probably lose) is still competitive.

Queensland: Ashgrove Still A Problem
The grass is green, the sun is shining, the lost PUPpies are coming home ... except in Ashgrove, where there are few of them, and therein lies the first-term Liberal National government's problem.  They're heading for an election they should win pretty easily, having led all term in aggregated polling and holding a huge majority, but their Premier is still at great risk of defeat in his own seat, is unpopular generally, and has no clear successor.   It's a bit of a mess and could make for a really bizarre campaign and election night sometime around March next year.

At the start of the month there was a ReachTEL that showed the LNP with a 41.2% primary, Labor with 36%, the Greens with 6%, PUP 9.5% and others 7.2.  This was converted to a 51:49% 2PP - my guesstimate is more like 52, but it's a moot point under optional preferential voting with known unknows in the party distribution anyway.  The important bits are that the LNP have a substantial primary lead on Labor and that the PUP vote was already crashing, down 5.9 points since early July.

The quarterly Newspoll confirms these trends.  The LNP primary is 39 (up 7), Labor 32 (-2), Greens an optimistic 10 (+2) and Others are down from 26 to 19.  Newspoll doesn't separate PUP in its published figures, but I'd be very surprised if most of that exodus wasn't PUP to LNP, in line with the national trend.  The published 2PP is 54:46, up five from a 49% that was a bit off beam last quarter, and the best result for Newman's government from anyone since Galaxy in May.  

At the moment, the scenario in which PUP wins many seats, potentially causing a hung parliament, is pretty much off the table.  On current figures they might only get a seat or two.  This is now shaping as a more or less straight LNP vs Labor election.  That's a very good thing for the LNP since the Labor primary continues to struggle.  A massive readjustment from the last election has always been expected, but there's still room for a pretty easy government win.  

But led by whom?  With monotonous regularity, polls continue to show Campbell Newman is struggling to hold Ashgrove.  The Together Queensland ReachTEL (union-sponsored) showing Newman losing to former member Kate Jones 58:42 did look a shade generous to Labor, but the 7 News ReachTEL of the same seat in the last few days, at 56:44, was not much better.  Indeed it was the same 2PP as found when Kate Jones was hypothetically included in the mix back in July.

(By the way, I have a pretty low opinion of Together Queensland's choice of poll questions.  Question 3 engages in blatant preamble-skewing by selectively giving respondents information about Campbell Newman's pay rises without mentioning their setting by an independent tribunal, or that the Opposition Leader also received large pay rises. The unstated insinuation that Newman gave himself a pay rise then pollutes  everything from question 3 on. In theory it shouldn't affect the first two questions, but if someone can't be trusted to word a question cleanly, should they be trusted to release all their polls rather than just the best ones?)

A positive spin was put on the small shift between these polls by the Courier Mail which reckoned Newman had begun to claw back lost ground.  However the TQ poll was the single worst published poll for Newman in Ashgrove in his Premiership, and polls of this size have margins of error of around 4%.  It's far from clear he was ever that far behind to begin with.

It is extremely rare for sitting Australian Premiers to lose their own seats, but I'm aware of no precedent for them to be in seats as hard to hold as Ashgrove in the first place.   A useful way to look at a seat like this is to track its lean compared with the state 2PP over time.  While 2PP is a messy concept on a statewide level in Queensland, Ashgrove was solidly Labor from 1992 to 2008, excepting a reasonably close result in 1995 (when the National/Liberal Coalition comfortably won the statewide 2PP but had to wait for a by-election to take power.)  In this time the seat has always been more pro-Labor than the election outcome, by margins ranging from three to nine points.  In 2009 it was so by 6.2 points, and in 2012 by 7.1 points.  

There is not going to be a normal sophomore effect in Ashgrove, because almost everyone knew who Campbell Newman was before the last election.  Plus he isn't popular within his own electorate, and he is running against the former member, whose profile is high and ratings excellent.  So all else being equal, Newman retains Ashgrove if the state 2PP is around 56-57% to the LNP.  Maybe. There are, furthermore, hints that not all is equal and that it's actually a bit worse for the Premier than that: voters outside south-east Queensland could well be coming home to the LNP after parking their votes with KAP or PUP, but in Ashgrove there is scarcely any vote for minor parties and this effect isn't being seen.

There are plenty of pieces about declaring it's just about all over for Newman unless he changes seat.  I think these declarations are a bit premature.  Firstly the polling since the election showing Newman so far behind all comes from a single pollster; seat polls are not that reliable so perhaps it is actually closer.  Secondly it's still quite possible the LNP will have a 55%+ 2PP win at the next election and that it will be large enough for Newman to just be carried over the line by that rising tide.  But these are not hopes I would recommend relying on.

While Newspoll did show slight improvements in Campbell Newman's net ratings, a -19 netsat (35-54) is still very bad.  Opposition Leader Annastacia Palaczszuk sits on exactly zero (36-36) and six points is a remarkably poor "better premier" lead for Newman given the size of his party's 2PP advantage.  ReachTEL ratings for Newman looked a little better given the tendency of the "Satisfactory" option to include a share of lukewarm positives, but I've been wondering if that is waning for both Newman and Abbott. My ReachTEL-to-Newspoll netsat conversion method (count half the Satisfactory reading as positive) hasn't been working as well as it used to lately.

NSW: Baird Maintaining Lead
I didn't previously cover the late August Galaxy that showed the Baird government cruising at 55:45 or the July-August Newspoll (54:46).  These were still very much honeymoon results for the new Premier whose Newspoll netsat stood at +26 (49:23).  The Galaxy poll question on corruption hit the nail on the head: although the recent corruption-related exodus to crossbenches and by-elections has been from the Coalition side, voters overwhelmingly (45-13) rate Baird better at tackling corruption than John Robertson, and even among Labor voters enthusiasm for their own leader in this regard is very weak (a mere 31:21 lead over Baird.)  This prompted Galaxy's David Briggs to declare that John Robertson "appears unelectable", a comment truncated by some reports that dispensed with the "appears".

Some gloss might come off Baird's significant early leads sooner or later but a recently reported 53:47 Essential result for September is again not that much evidence of that.  Again the reporting draws comparisons with a "March" Essential result when in fact there is a full run of published Essential figures from November to May: (2PPs for Coalition) 60, 57, 58, 53, 57, 54, 54 (O'Farrell resigns during April), 51.  Looking across an incomplete picture of polling from all pollsters earlier in the year it is hard to say if the government is doing better or worse now than just before O'Farrell quit, but there's not a great deal in it.  What I don't doubt is that but for the honeymoon period for the new Premier, things would now be closer.  All else being equal, they should be expected to become closer prior to the election.  All the same it would be hard to find serious observers who take the idea of Labor winning the election to be held on March 28 all that seriously.

SA: Dismal Days For Opposition
Still licking their wounds after an election defeat followed by the defection of Martin Hamilton-Smith, the Liberal Opposition led by Steven Marshall is further embarrassed by a three-month Newspoll sample showing it trailing.  (And in South Australia, a 49% 2PP for the Liberals on current boundaries would lose them about another five of the 21 seats they now hold in the 47-seat parliament.  The boundaries may well get massively changed soon, but that's another story.)  

It's normal to see a victory bounce for a re-elected government in its first six months post an election, so it's not surprising to see Labor up on the 47% 2PP they won with back in March. However the term "defeat dip" (used by a Pollbludger poster) is a better one for what is happening here, since the rise on the 2PP for Labor comes mainly because the Liberals have lost a lot more votes to Others than Labor has.  Accounting for sample error, Labor may not have lost any primary votes at all, but there's nothing to show their primary vote increasing on the election result.  If another election were held it likely would increase, since many of those picking Others are probably disgruntled ex-Liberal supporters who might preference or even vote Labor.  The massive 21% for Others can't really be taken on face value.  It's effectively the highest in SA ever by some margin, although a 24% reading was recorded in one 1998 poll that counted everyone but Labor and Coalition, including the then very popular Democrats, as "Others".

Meanwhile Premier Weatherill gets a victory bounce in his personal netsat (up from zero before the election to +8), Steven Marshall's is unchanged, and Weatherill's Better Premier lead of 15 points is consistent with the Others preferences breaking fairly evenly between the majors.

Update: Nationwide Morgan State Polls
A good laugh can be had at my plans in writing this article since less than an hour after its release Morgan unleashed state polling of every Australian state.  Perhaps I should now make this article "Recent Polling in Six States", but in reality the WA and Tasmanian Morgan samples are a little bit small to get excited about.

The Morgan polling is being marketed as superior to Newspoll, since it is based on one weekend's data rather than three months of rolling averages and since it breaks out the PUP total.  There has not been much experience with purely SMS-based polling of voting intention but what there has been (Morgan's SMS polls very close to the election in 2010 and exit polling in 2013) has been fairly promising.  I would also say on the basis of these results that the method doesn't seem to show the skew to Labor of the pollster's regular multi-mode series that includes a face-to-face component.  The results do look a tad Greens-friendly, but those excluding Tasmania are no more so than recent Newspolls on average.

However the published 2PP for NSW (53:47 to Coalition) is not accurate given that the published primaries are 46% to the Coalition and 32.5% to Labor. In 2014 a Coalition primary lead of 51:25.6 turned into a 64:36 2PP.  It's likely preference flows from any given party would be slightly stronger to Labor now, and also the poll has the Greens up slightly, but these differences are just not going to be worth an 11% 2PP swing off a primary vote swing worth six points.  I think a correct 2PP for these primaries under the NSW optional preferential voting system is more like 55-56%.  I don't know enough about their methods with these SMS polls to say what they are doing wrong here (though if these polls use respondent preferences, that would explain it - but so would forgetting about OPV).

A similar concern applies in Queensland where the published 42-35.5 primary lead for the LNP is supposed to pan out to just a 51:49 2PP.  There is much less scope for rounding to cause differences given that Morgan publishes primaries to half a point, and I think that given the use of the OPV system, 52:48 is more accurate.  I do not agree that an election with the published primaries would be too close to call.  It's very, very likely that the government would win.

I don't, on the other hand, have any issue with the 54:46 2PP published from the given primaries for Victoria (though the Greens vote of 18 seems a bit high). So the net effect of these polls should be to slightly improve Labor's estimated position in Victoria (the primaries from these polls are more useful than Essential's given the larger sample size), improve the Coalition's in NSW, and make very little difference in Queensland.

The small state samples in SA and WA are both more Coalition-friendly than recent Newspolls, but the sample sizes (around 500 each) are not very reliable.  The Liberals in SA might have improved their primary recently, or it might be just a noisy sample.  Still, at over half the size of the Newspoll and more recent, the Morgan data shouldn't be ignored.  Especially, it's useful to note that some of the Others vote in SA appears to be going to PUP (and I find Morgan's measure of combined others - 14.5% for PUP+Family First+ Ind/Others) to be likely closer to the mark than the most recent Newspoll's).

The Tasmanian sample with only 293 respondents and claimed percentages of Liberal 39.5% Labor 33.5% Greens 19.5% PUP 3% Others 4.5% is not consistent with the far larger sample ReachTEL taken a few weeks ago, and  I somehow doubt the Liberals have lost eight or nine points in that time.  Even though the small sample size implies an error margin of up to 6%, I suspect this sample is further out from the truth than that and I have had concerns about the reliability of Morgan's Tasmanian samples for a while.  It's possible I'll eventually aggregate this result with a very low weighting but I would like to see another by the same method first to benchmark possible house effects of this method in the state.

The Morgan better premier results have no undecided figures, and hence look a bit odd.  They mostly seem to lack the innate advantage to incumbents normally seen in polling of this type.  If they are asked as follow-up questions after the main question, it would be useful to know what percentage of respondents actually answer them.  They might be catching only the most strongly-held views.

I've asked Morgan to explain their NSW 2PP figure methods and will report any useful response.

Update 13 Oct:  We do finally have something more meaningful in WA with a quarterly Newspoll showing a 50:50 2PP.  Barnett records another new worst netsat as WA Premier (-24) while McGowan has a +18 netsat and a 41:38 better premier lead.

Queensland ReachTEL (20 Oct): A ReachTEL released for Queensland on 13 Oct (taken on the night of 9 Oct) escaped my attention.  This time I agree with the 2PP of 51:49 off the primary figures given, which are mainly notable for the PUP slide continuing to a mere 7.2%.  Newman's performance ratings show slight improvement.  South-East Queenslanders support daylight saving (55.7-38.9) while Queensland voters outside SEQ do not (20-75.5).  An interesting question is "Do you think the Palmer United Party will win any seats in the upcoming Queensland state election?" PUP supporters almost unanimously "think" (translate: hope) they will while LNP supporters are the most sceptical (31:69).  Labor and Greens supporters are both around 60:40 in thinking at least one PUP will get up.  If PUP doesn't lift from its current support base there is a real risk of them missing out entirely, especially as they now have no remaining incumbents.

5 comments:

  1. Kevin,
    I live adjacent to Ringwood State electorate in Victoria, and it's certainly evident that both parties believe that it is in play. Unlike everywhere else in the eastern suburbs, there are masses of posters of the Liberal candidate (who is currently Member for the abolished seat of Mitcham). At the same time, the Labor candidate has opened a campaign office three months prior to election day. The swing required is a substantial 6.3%, compared to the neighbouring seat of Bayswater 6.8%.
    On the face of it this might be misleading. The new Ringwood has been carved out of Mitcham and Warrandyte. The Liberals won Mitcham last time unseating a Minister in the Bracks and Brumby Governments. That meant that Labor ran a serious campaign in that electorate, whereas they were almost invisible in the rest of Ringwood, and in most of the neighbouring electorates. That suggests to me that the Bayswater margin is inflated and that relatively speaking the margin in Ringwood might be under-stated. That is of course trumped by the implications that internal polling suggests it's contestible.

    ReplyDelete
    Replies
    1. Thankyou! Local insights much appreciated.

      Delete
    2. I also think outsider caution about Morwell is misplaced. Morwell is a very working-class area and had been safe Labor for donkeys years before neglect, a strong Nationals campaign, and a hopeless local member saw them grab the seat a few years back. The Nationals have benefited from incumbency and have done well to hold on as well as they have (in comparison to the Liberals in the similar Hunter region in NSW), but for the first time they've got local issues running against them and I'd be really surprised if it wasn't in play on election night.

      Delete
  2. I presume there is no NT polling? It would be interesting to see how PUP is going there, given that they have a few state MPs. However, it is obviously a hard region to poll.

    ReplyDelete
    Replies
    1. The only NT polling I know of since the last election is a Telereach poll of Greater Darwin in April:

      http://www.ntnews.com.au/news/northern-territory/support-is-waning-for-clp-poll/story-fnk0b1zt-1226889898956

      I am not sure how reliable Telereach are. However (i) that poll predates the defection of three CLP MPs to PUP and the subsequent return of one of them to the CLP (PUP only have two there now) (ii) there has been no polling of the outlying regions and I don't think there ever is any, even at election time.

      Delete