Tuesday, August 22, 2017

Poll Roundup: Newspoll Moves! It Is Alive! (includes same-sex marriage polling)

2PP: 53.5 to Labor (+0.3 since last week)
Labor would win an election "held now" with a large majority

Time for another roundup of the state of play in federal polling.  This week's exciting development was that Newspoll, after a record six consecutive 2PPs of 53-47 to Labor, finally moved by a whole point to 54-46!  I am especially excited by this because had this week's result been another 53-47, I was going to lead off with a mock Death Certificate, and now I do not have to.

This briefly took my aggregate to an equal term high 53.8 to ALP (I tweeted this as 53.9 but later found a very small error) but then Essential moved back a point the other way after two 54s, and the primaries in YouGov didn't do a lot, so I currently have it at 53.5 to ALP. That was the Coalition's winning lead in the 2013 election, and also just shy of what Labor had when Tony Abbott was removed.  

The strange 2PP results of YouGov continue to baffle - this week it has the Coalition with a 51-49 lead off primaries that would normally imply something like 53.1% to Labor.  I discussed this issue last time, and the average difference between YouGov's 2PPs and the last-election 2PPs for their primaries is now running at a massive 2.9 points.  As noted last time, it's possible that the current last-election 2PPs are overstating Labor's lead, especially because of One Nation issues, but it's highly unlikely that they're doing so by three points.  Much more detail on the YouGov mystery from William Bowe here.

Once again an event of great interest to the political beltway - the Section 44 eligibility crisis - was widely expected to produce massive polling movement but so far hasn't amounted to much.  Here's the smoothed tracking graph:

Prime Minister Turnbull's self-imposed "Newspoll clock" reached eighteen consecutive Newspoll losses.  As noted last time, this was an unfortunate choice of metric by Turnbull when he was a leadership aspirant, because the new Newspoll is less bouncy than the old one.  Historically, the old one would have given him almost a 50-50 chance of getting at least one 50-50 reading sometime in the next twelve polls.  Based on the properties of the new Newspoll, it might be more like 25%.  

But sometimes things do happen that move things sharply in a government's favour.  There are several cases of governments in aggregated polling positions as bad as this one's or worse recovering to rough equality within the span of a dozen Newspolls.  This happened with the Keating government in late 1992, early 1994 and late 1994, and the Howard government in early 1998, mid-2001 and early 2006.  (Later such recoveries came at the cost of incumbent Prime Ministers.) Of these, only the Keating government in the early 1994 case had been polling as badly for almost as long as Turnbull's government has, but so few governments have polled so badly for so long without something changing that that doesn't really tell us much.


This week's Newspoll netsats went back to where they were four weeks ago, with both leaders on a net rating of -20, after Turnbull scored a post-election high of -12 and Shorten -15 the previous fortnight.  I suspected that last fortnight's gains were a temporary bounce arising from the announced foiling of an alleged terrorist plot to blow up an Etihad jet.  This is only the ninth case in Newspoll history in which a PM has gained eight netsat points or more in a poll and then lost all the points thus gained (or more) next time.

Essential two weeks ago had Turnbull on -8 (his best since last November) but this was at the same time as the temporarily not-so-bad Newspoll.  It also had Bill Shorten on -7 (his best since January, and ditto).

Turnbull jumped to a term-high better PM score of 46% in last fortnight's Newspoll, and an equal-term-high lead of 15 points over Bill Shorten, but that's all history too, with Turnbull back to a 43-33 lead this week.  Again, this is still pretty good given that his government is trailing so badly.

David Crowe in The Australian had a very interesting suggestion (Newspoll: Turnbull’s house tactics turning voters off) that over the last year or so things have tended to go downhill for the Coalition and Turnbull when Parliament sits.  The sample size is obviously nowhere near proving causation, but it was very obvious in the last session especially which side had its act together tactically and which did not.

Same-Sex Marriage Polling

We have now seen the three most regular pollsters all move into polling on the result of the same-sex marriage postal ballot, which will be held later this year unless struck down in High Court hearings in September.  See Newspoll, YouGov and Essential (1), Essential (2), Essential (3), Essential (4).

In my initial piece on the concept of a postal plebiscite I mentioned that the evidence on anti-SSM voters being more committed than pro-SSM voters was unclear and tended to vary a lot between surveys.  In this week's surveys, the polls suggest that voters who support same-sex marriage are more likely to take part than those who oppose it:

* YouGov finds a 35-point lead for Yes among "very likely" respondents but only 15 points for others.

* Essential finds 92% of supporters saying they will participate with 4% saying they won't, compared to 74-19 for non-supporters.  

* Newspoll finds those who say they will definitely vote breaking 67-31 to Yes.  This is actually about the same proportionally as Newspoll's overall 63-30 result.  (The figure for those who say they won't vote should be ignored, as they were only 3% of the sample).

Newspoll finds a 49-43 response to the question "Are you in favour or opposed to the postal plebiscite?" while Essential finds a 39-49 response to "Do you approve or disapprove of the Government’s decision to hold a national voluntary postal vote on the issue of same-sex marriage?" It's possible Newspoll's wording could have confused some respondents who might have thought "in favour" meant supporting same-sex marriage.  On the other hand, Essential's question by mentioning that it is a Government decision may connect responses to impressions of the Government.  Anyway there's not a great deal in it between these responses, and we can conclude that public opinion about the holding of the postal survey is somewhere around evenly split.

If the patterns suggested in the polls hold true then Yes would seem to be set for an overwhelming victory.  Yes has 68% of the two-answer vote in Newspoll (63-30), 64% in Essential (57-32) and 64% in YouGov (59-33).  With the polling evidence of Yes supporters being more likely to participate than No voters, it would seem a Yes response of 70% or more was possible.

Perhaps it is, but overseas experience of polling for such votes should make us rather cautious.  Young voters are more likely to not be enrolled even if they believe that they are, or to have their form sent to an incorrect address and not reach them in time.  Final polls before the Irish referendum all overestimated Yes on a two-answer basis, and did so by an average of nine points.  I remembered something similar happening on US state ballot measures dealing with same-sex marriage, but on checking those for which I could find data I found the average error to be only around three points on a two-answer basis.  (Typically this results from an underestimate of opposition.)  Commonly, this pattern of underestimating opposition to same-sex marriage has been explained as caused by "social desirability bias" (respondents are shy about admitting to opposing same-sex marriage in case the interviewer is offended) but this explanation is sometimes disputed.

It would be hard to find any precedent for a postal vote survey on same-sex marriage.  Among other things, the dispersed timing of voting may make it more difficult for intense opposition to be orchestrated through religious gatherings, rallies and media focus.  I suspect that the projected turnouts from these polls (which can be taken as implying a turnout in at least the 70s and possibly the 80s) are a bit on the generous side.  Those who refuse to be involved in polls of any kind are probably disproportionately unlikely to vote, and poll respondents often over-report participation generally (such as having voted in past voluntary elections).   

There was a Newspoll question showing high support for "guarantees in law for freedom of conscience, belief and religion" if same-sex marriage is allowed.  For all the interest in exemptions on the pages of The Australian, this is a niche issue, and I suspect that without those guarantees being specified, the response is just an intuitive "sounds good".  Specific examples might not be so popular.

Other polling

As usual there are many other issue polls available. Adrian Beaumont has a summary of lots of them here.  Among the newest ones, YouGov finds 45% saying Barnaby Joyce should step aside while his citizenship is cleared up to 38% saying he shouldn't.  Essential finds voters evenly split (35-38) on Australia assisting the US in a war with North Korea ("depends who started the bloody thing" is not available as an option).  Essential also finds voters of all parties (61-22) believing decisions to go to war should be made by Parliament not the Prime Minister.

The Fifty Acres press release re YouGov made the ridiculous howler of claiming "The majority of Australians think Deputy Prime Minister Barnaby Joyce should step aside [..]" when 45% is not in fact a majority.  I recently gave the Bob Brown Foundation a Porcupine Fish award for doing exactly the same thing.  The BBF has a long record of form in this regard but on the hand as an outlet for the findings of an actual pollster Fifty Acres should surely know much better.  So ...

Wirrah Award For Fishy Poll Reporting
Awarded to 45 (hey that's 50 isn't it) Acres

ReachTEL update (Thursday)

A Sky News ReachTEL has been released. The headline 2PP is 52-48 by respondent preferences but I get 54.4% by last-election preferences so this is actually a rather nasty one for the Coalition and their worst set of ReachTEL primaries in this term.

Will One Nation really poll 9-10% at the next election with 75% of that going to the Coalition?  Again, if so then the next election is closer than it looks, but such preference shift patterns have fallen over by election day often enough before.

Anyway for the meantime adding this ReachTEL to my aggregate has pushed it up to 53.8 to Labor.

Meanwhile I'm adding some more sidebar fun - a Section 44 not-a-poll.


  1. Hi Kevin, at the risk of exposing my Ignorance of Irish affairs, were you referring to the Scottish or Irish referendum vote in the 4th last para above?


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.