Monday, April 13, 2015

Poll Roundup: Discordant Polls But No Real Change

2PP Aggregate: 52.5 to ALP (+0.1 since last week)
ALP would win election held now with small to moderate majority

As my colleague Mark the Ballot notes, federal polling lately has been decidedly two-tone.  With the exception of Essential (which Mark ignores), once you adjust recent polls for what is known about their house effects, they tend to be either reasonable for the Coalition (the last two Newspolls, the most recent Morgan) or very bad for them (ReachTEL, Ipsos).  Individual poll results that are strong or weak create excited commentary in the media but on average there is not a great deal happening.  Here's the smoothed tracking graph:

There's a weak hint of a slow move to the Coalition over recent weeks, but there's been no significant 2PP change for over a month.  Indeed, because the aggregate had some lag from the Coalition's dire February polling for a few weeks, it may be that nothing has really happened in 2PP voting intention since the beginning of March.

I was going to emphasise these results as a classic case of "house pollster bias", in which news outlets interpret the political world according to the latest bounce in the outlet they commission.  There has been quite a bit of that about, but it's notable that, for instance, the Australian reported the Ipsos poll in the same article as Newspoll.  Also The Australian has recently reported Morgan poll results online (though anti-Murdoch Newspoll conspiracy theorists ignore that and are blind to all other evidence that contradicts their sledging.)  There's still a long way to go but I do think the Australian media are getting slightly better at trying to cover the cross-polling picture.

Recent Polls

In the last instalment a few weeks ago I noted that when the government rallies as sharply as it had done in recent Newspolls, this is usually followed by a drop in the Newspoll after.  Surprisingly this time this did not happen, with a second consecutive 51:49 (and off slightly better primaries for the Coalition than the last one).

Newspoll continued to show improvement in Tony Abbott's personal ratings, with Abbott's netsat up to a still very bad -26, just one point shy of where it finished 2014.  The suggestion from Newspoll is that Abbott's personal ratings have just about recovered the damage done by the Prince Philip knighting, the loss of the Queensland election and the subsequent spill attempt, but his own limited standing has recovered much more slowly than his Government's.  The more notable Newspoll netsat was Bill Shorten's, -18, his worst to date, including his highest dissatisfaction rating so far (51%).

Shorten clung to a one-point lead as "better Prime Minister" - bearing in mind that any lead for an Opposition Leader on this indicator is good, as it skews strongly to incumbent PMs.  Shorten is only the third Opposition Leader of the Newspoll era to lead as better PM while having a netsat as bad as this or worse.  The previous cases involved John Hewson vs Paul Keating (six times in late 1993 when Keating was hated because of a horror budget while Hewson was discredited by an election loss) and Tony Abbott vs Julia Gillard (thirteen times between August 2011 and April 2013).  Both those movies ended in Opposition victory - though in one case under a different leader, and in the other against a different Prime Minister.

Ipsos had the same netsat for Tony Abbott as Newspoll, but a much milder -2 for Shorten, who it also had still with a healthy 8-point lead as Preferred Prime Minister.  Usually Ipsos' leader results seem to be less scathing than Newspoll's (as was also true of Nielsen before Ipsos) and in this case the lack of difference for Abbott but large difference for Shorten is likely to be down to Ipsos having a more ALP-friendly sample than Newspoll this time around.

Indeed, the Ipsos result of 54:46 to Labor by last-election preferences (55:45 respondent-allocated) was quite a surprise because this poll has hitherto appeared to lean to the Coalition both at state and federal level.  (Even with this result I have its average lean so far at 0.9 of a point.) Ipsos often produces rather high ratings for the Greens and this was no exception, giving them a primary of 13%, compared to an average of 11.25 from other pollsters (an average which based on past election trends is probably still overstating things, although there may be some bounce from the Greens' win of three seats in the NSW Lower House).

Joe Hockey is reported as having a rather bad net approval rating (which Mark Kenny mistakenly calls an approval rating) of -25. Ipsos also finds decreasing voter resistance to raising the GST, although voters are still against such a raise 59:37.  41% of voters nominated the Coalition as best to handle the economy compared to 32% for Labor - an underwhelming lead for an incumbent conservative government.  The poll also quite unusually included the Greens as an option for that question, and needn't have bothered as they only polled 3%.

This followed on from a similarly bleak ReachTEL for the Coalition before Easter (also 54:46). The ReachTEL, the first in nearly two months, also showed modest improvement in Tony Abbott's ratings since early February, with an insignificant negative move in Bill Shorten's. The ReachTEL showed continuing but slightly weakening support for Malcolm Turnbull or Julie Bishop to be Prime Minister instead of Abbott.  However, 52.6% of Coalition supporters wished to retain Abbott compared to 24% for Turnbull and 21.5% for Bishop, a much larger lead than in similar polls earlier in the year.

The ReachTEL showed that voters generally prefer cutting government spending as a way to reduce national debt, compared to raising taxes or selling assets.  Labor voters are more divided than Coalition, PUP and Others supporters, and Greens supporters prefer raising taxes.  (As Ipsos showed, Green and Labor voters aren't that convinced that reducing debt now is a high priority anyway.)

Last week's Morgan was odd: off rather strong primaries for the Coalition (40.5 to 36 with 12.5 for the Greens and 11 for others) it came out with a last-election preference figure of 53:47 to ALP - normally these primaries would have produced a 2PP of 51.5 or 52.  Because Morgan releases 2PPs rounded to a half, my aggregate only modifies the released 2PP very slightly in such cases.  Given that Morgan federal polls lean to the ALP, it was actually quite a strong poll for the Coalition.

And finally Essential produced 53:47s the last two weeks in a row; this week's figure has not been released yet and I will update this sometime tomorrow night after it has.

These polls bring up an unwanted anniversary for the Coalition - it is now more than a year since it has led the 2PP in any published poll.  In a few more days the Coalition will not even have tied the 2PP in any published poll over 12 months, though one Morgan in October becomes a tie when adjusted for house effect.

United Voice Seat Polls

Some sketchy reporting is available of some United Voice seat ReachTELs said to show Labor ahead 57-43 in Eden-Monaro (NSW), 54.7-45.3 in Swan (WA), and it seems winning (some ambiguity in the reporting) in Bonner (Qld) and Hindmarsh (SA).  However this is yet another case where a union-commissioned poll uses the preferencing method that happens to give the ALP the biggest lead (respondent-allocated, at least in the case of Swan).  I don't have enough detail on the Swan poll to know if the primary vote totals include undecided voters or not.  If they don't, then the last-election preferences would be about three points less generous to Labor using national results, or close to four points less generous using results for this specific electorate.  Which comes out to about what should be expected, given the apparently large swing against the government in WA, which is now the Coalition's worst state in terms of swing since the last election.

If there is an undecided portion included (meaning that the Greens' share of the non-major party vote increases) then the last-election 2PP might be a bit better than the state swing, but I doubt the value of trying to work out how much.  Publicly released seat polls in Australia lately are really struggling for legitimacy as reliable indicators of anything we couldn't tell from national and state pictures, and that's even when we have all the data and no reason to suspect selective release.  Trying to make sense out of selective excerpts from sources that haven't published anything near full data, as any kind of evidence of anything, is hardly worth it anymore.

Queensland State Poll

We do have the first high-quality polling in Queensland, where the new Palaszscuk government confronted an immediate crisis upon finding that new MP Billy Gordon was facing domestic abuse allegations and had not disclosed very old juvenile criminal records during preselection.  Palaszczuk has had Gordon thrown out of the parliamentary Labor Party and he sits as an independent but has said he will support Labor on confidence and supply.

A new Galaxy poll has been released with a 2PP outcome of 50:50, but it uses a combination of 2012 and 2015 preferences, rather than just preferences from the 2015 election.  This unusual move may be an attempt to pre-empt a less strong preference flow to Labor at the next election than the anti-Newman flow seen in 2015.  Whether there are any signs that preference flow is easing already is unknown to me, but I would be surprised if there was a big shift at this stage.  Anyway, by 2015 preferences Labor would be ahead around 52:48, so this is not a bad poll for Labor given the Gordon crisis.  The new Premier has started with very positive ratings (53-24).  Voters predictably think Billy Gordon should resign from parliament (62:26) but on that score the only one whose opinion really matters is Billy Gordon.  I doubt he's going anywhere; it pays well.

(There were earlier some Morgan SMS polls showing LNP leads of 51:49 but after that poll recorded another poor final result in NSW I am treating it with a great degree of caution.)

I am writing something refuting parts of a piece in The Australian that claimed that Greens voters tended to be rich atheists, but I think I'll post that as a separate article to make it easier for people to find it.

Essential (Tuesday 14th)

Essential this week came in at 52:48 to Labor, and it's pretty much dead on that mark by the primaries, so my aggregate moved down 0.1 of a point to 52.5. (I've updated the header, but not the graph.)

Essential concurred with Newspoll in finding Bill Shorten on a career-worse netsat, in this case -9 (33-42).  Tony Abbott stayed at his second-worst level as PM, of -25 (33-58).  Shorten led Abbott by three points as preferred PM (35-32).

Oddly, Essential's sample has switched from leaning towards thinking the government is too harsh on asylum seekers to leaning slightly towards thinking it is too soft.  This although respondents are less likely to believe all asylum seekers arriving by boat should be sent back.

Essential also had an interesting question on MPs who quit their parties, with Liberal voters generally likely to support such an MP facing an election for the seat.  That's quite impractical for the Senate since a minor party Senator's seat would be won by a major party, quite aside from a state by-election for a Senate seat being an expensive exercise, so I suggest that 52% of Liberal voters, 36% of Greens voters, 34% of Labor voters and 40% of Others voters haven't really thought this through.  Just about any sanction that singled out a voluntarily leaving MP would simply cause dissident MPs to act the goat to get themselves booted out.


I forgot to mention that a deadline has passed in my sidebar Not-A-Poll! 13.9% of voters (at that time) incorrectly predicted Tony Abbott would be gone by the end of March.  17.5% out of 360 votes now have him winning the next election, and another 15.8% now think he will make it to that election but lose it.  Departure by the end of June remains the most popular option with 26.1%.


  1. Well it now makes a change to be 'rich atheists' from 'dole-bludging, dope smokers' 'The Australian' and News Corporation got into a rage over Christine Milne's questioning in regard to the paying of corporate taxes. It is difficult for many to realise that inner city voters are mostly social progressives and not religious conservatives. Especially, those of us babyboomers that were forced to go to Sunday School and now couldn't care less about any religious organisation. To get the 'so-called' atheist demographic figures 'The Australian' must have used the census figures but that surely would reveal the 'no religion' figures only. What a joke. Are all Green voters 'atheists'? How the hell do they know that? I am sure there would be plenty of Labor and Coalition voters who have 'no religion' as well. Isn't that demographic getting towards 1 out of 4 of the population and growing.

  2. The NSW census data breakdown gives a total for "persons of no religion". They define this as "Persons who stipulated on the census form: 'No religion no further details', 'Agnosticism', 'Atheism', 'Humanism' and 'Rationalism'". I can remember seeing at least some of these options included on the census last time though I don't recall the exact form of the question.

    Newtown has the highest percentage of "no religion" at 39.3%. Nationally "no religion" is 22.3% and growing but in NSW it is only 17.9%.

    1. Thanks Kevin. Obviously, the 'old churches' still have an impact in NSW.

    2. I did wonder if it was just that NSW had more people adhering to non-Christian religions because of its more multicultural population, but it seems it does have the highest proportion of Christians as well as a low proportion of "no religion".

  3. I think Vote Compass had some interesting data on the demographics of Green electorates, and where atheists live.