Thursday, July 9, 2015

Poll Roundup: Abbott, Shorten Racing To The Bottom

2PP Aggregate: 52.5 to Labor (+0.2 in a week, highest for ten weeks)
Labor would almost certainly win election "held now" with small to modest majority

This week has seen the debut of the new Galaxy-run Newspoll, a hybrid robopoll and online poll with a sample size of over 1600.  There's also been another Fairfax-Ipsos, the usual Essential and far too many dodgy skew-polls, so there's more than enough for another poll roundup.

Not much changed in as week in the media issues mix, with the government sinking further into the mire of internal confusion on both same-sex marriage and, of all things, Q+A.  The week has seen an open clash between frontbenchers Abetz and Pyne over Abetz's controversial same-sex marriage comments (for my own, entirely critical, view of which see here).  

Same-sex marriage is not an issue that normally drives voting intention, but it may be one that discourages people from changing their vote when they otherwise would have done so (for "Republicans" there read "Liberals" here in Harry Enten's take on the US situation)  . It's also one that can cause damaging public schisms on both sides of the chamber.  The PM's own tactics on the issue are mysterious - it seems he will at least go through the motions of opposing it, but opinions vary wildly as to whether he is doing his utmost to prevent it from passing, or trying to secretly engineer its passage.  Some within the Coalition are totally fuming about the latter possibility, but as Paula Matthewson points out, if it is true, what can they do?  

The Coalition could look back to a previous major gay-rights issue for a hint what not to do.  In 1994 Alexander Downer sacked John Hewson from his frontbench for agreeing with Labor's policy on neutralising Tasmania's anti-gay sex laws.  This caused a 28 point hit to Downer's already damaged netsat, from which he never recovered.  Months later most of the Coalition voted with Labor anyway.

Tony Abbott and his team would be pleased that the Royal Commission into union corruption they started will be giving them some different headlines for a while, whether it causes serious harm to Opposition Leader Bill Shorten or not.

This week's polls 

This week Ipsos came out with another 53:47 to Labor, while both Essential and the new Newspoll had 52:48.  However, the Newspoll result was off primaries that, all else being equal, would slightly more often have produced a 2PP of 53, so the 2PP was probably rounded down.

My aggregate is treating Newspoll as a new poll by an established pollster, so it is included immediately with an assumed house effect of zero.  Over coming weeks we will get a clearer picture of how accurate that is, but so far there's no reason to do otherwise.  Also the Ipsos result has weakened evidence for Ipsos having a strong house effect for the Coalition.  My estimate of the average lean of all Ipsos polls so far is down to about 0.6 points, and my trending estimate is down to about 0.4.  As a result my methods no longer treat Ipsos as having a specific house effect, but Labor now gets a 0.1 point global correction.

Taking the primaries into account I aggregate the Ipsos as 53.1 to Labor, the Newspoll as 52.3 and Essential as 52.1 and the new reading is 52.5, Labor's highest for ten weeks (in which the only movement really was a short-lived and small Budget bounce.)  Here's the smoothed tracking graph:

The Greens recorded 16% in Ipsos, which is their equal second-highest reading from any national pollster ever (the highest was an almost certainly rogue, or at best a very ephemeral spike, 17 from Nielsen last year).  This continued their surge noted last week, but Essential wasn't having a bar of it.  William Bowe covered this at Crikey (it's paywalled but you can see the spiffy graph of the Greens vote by pollster for free).  

I think it's very likely that differences in response between online polling and all forms of phone polling are indeed the main explanation here, but I have a specific suggestion as to why.  An online poll has to offer "don't know" as an upfront option on voting intention while a phone pollster doesn't need to include it in the readout, and I don't believe ReachTEL allows a don't know option on its federal polling (it often allows "undecided" for local commission polls).   I think voters sometimes pick Greens as a form of glorified or protest "don't know" response during times of disgust with both major parties, and that this is a major reason why most polls overstate their vote (with Essential one of the less prone to this).  In this case, online polling methods may be an advantage.  


Any historic comparisons involving new Newspoll data will forever more come with an asterisk, because of the radical methods changes that have been made.  Fortunately the question design of the old poll has been preserved, but robopolling and online polling are both quite different to landline polling (especially in that both lack the aspect of interaction with another person), and it remains to be seen how this will affect leadership ratings.  Of course, this is not the first time Newspoll has made methods changes, but those made in the past would not have affected the basic mode of surveying.

So far Essential's online-polling leader ratings have been quite benign compared to others, but the first Newspoll shows no sign of that, apparently picking up where the old one left off.  Bill Shorten records a "new low" of -28 (28-56) and Tony Abbott is down five to -27 (33-60), his "worst" for three and a half months.  Their combined tally of -55 is lower than any poll except the one in late February (when Abbott's -43 was the major culprit).  Where this one really stands out, assuming the methods differences are not to blame (something it may take years to be sure about) is in the extent to which both leaders are disliked at the same time.  In the whole history of "old Newspoll", there were only four polls in which both leaders were at -27 or worse: one in December 1994 (Keating -27 Downer -45), and three in May - July 2012 (Gillard vs Abbott; -30 vs -29, -29 vs -27, -33 vs -31). 

Ipsos had Abbott down nine points to -23 and Shorten down 14 to -20, having not recorded the severity of the Shorten plunge as steeply as other pollsters last time.  Ipsos had Shorten continuing to lead as preferred Prime Minister (43-39) while Newspoll had a tie (39-39).  The usual caveat implies: Opposition Leaders usually only break even on this score when the Opposition is about 53:47 up, so Shorten is actually punching slightly above his party's weight here - or perhaps more accurately, Abbott is underperforming.  As ReachTEL showed last week, outside the Coalition's support base there is very little liking for the PM at all.

Ipsos also had leader attribute polling (see tables) including comparison tables with readings in November.  Honours (or dishonours) are pretty evenly split between the two here, with Shorten thrashing Abbott by 34 points on "open to ideas" and 29 points on "has a firm grasp of social policy".  On the other hand Abbott is ahead by 18 points on resisting minority-group influence and 20 points on the "ability to make things happen".  

In Abbott's case, more recent results than November are available as Ipsos polled ten of the same questions in late February.  Since that nadir, Abbott is up 30 points on having the confidence of his party, and up six to twelve points on competence, vision, economic policy, strength and making things happen.  However, he's made little progress on social policy and minority group influence (though in the latter case his score wasn't a worry to begin with) and on trustworthiness and openness to ideas he is actually down a point.

A lot of Opposition Leaders don't actually last as long as Bill Shorten has without being elected, resigning or being dumped; he's a few months shy of the median for new Opposition Leaders who have not held the position before.  Historically his current ratings after nearly two years are about as bad as Howard's (first go) and Crean's; at the same stage Abbott and Snedden were also polling poor personal ratings, while Hewson, Beazley, Hayden at least were all popular.

Polling Re The ABC

Naturally there has been further polling on ABC bias.  Essential this week found that 22% of respondents including 42% of Coalition voters think the ABC is biased to the Left, 3% think it is biased to the Right, and 36% think it is unbiased; a massive 40% say they don't know, with Labor supporters especially vague.   This is actually one of the less flattering results the ABC has polled in such surveys, both in terms of the proportion of decided respondents finding the ABC unbiased (59% cf 65% in Nielsen 2014) and the ratio of perceptions of left-bias to right-bias (Taverner 2013 has 17 left and 5 right).  Most likely a little blowback from the Zaky Mallah incident here.

An intriguing commissioned ReachTEL (PDF) was issued by left-wing thinktank The Australia Institute.  The poll polled Liberal seats North Sydney (Hockey), Sturt (Pyne) and Wentworth (Turnbull) finding over 50% opposition to the reduction of ABC funding in last year's budget.  Opposition among Labor and Green supporters was predictably almost universal in the two NSW seats, but slightly weaker in Sturt.  

The poll also asked "Would you support or oppose including the functions of the ABC in the Constitution to protect it from political interference?"  This proposal had support levels well over 50% and in the case of Wentworth over 60%.  In Sturt even 50% of Liberal supporters approved and in the other two about as many (c. 40%) approved as disapproved. 

The second finding was uncritically reported as fact by a range of news sources, in a typical example of the synergy between activists and the press in which the activist gives the journo a ready-made polling story and gets back a cushy write-up.  In this case the first question to be asked was this: The Australia Institute's report released Questions 2 (support for funding cuts) and 5 (support for constitutional change to protect the ABC) of its poll.

Question 1 was very likely voting intention, but what were questions 3 and 4?  Without knowing what they were, I cannot rule out that they might have primed the respondent to answer Question 5 in the rather startling way it was answered. I asked TAI about this on Twitter and initially reported that I hadn't received a reply, but one was actually sent by the time I had finished and posted the article, and I've also since had an email from them. (My apologies for implying they were slack getting back to me - their reply just wasn't picked up in my notifications stream.)  They say that all six questions will be in the "larger report on the ABC due out soon" and also that while the intermediate questions were ABC-related they do not believe they would have influenced the answers to Question 5.  We'll be able to see soon enough.

The second is whether the yes response to Question 5 really says more about the reaction of voters to completely novel suggestions in polls than whether the idea of using the Constitution to "protect" the ABC is a good idea.  It is hardly something the vast majority of respondents would have ever even thought about before.  I think this might be what some UK poll-watchers like to call a pony poll: an idea that sounds fabulous for the seconds it takes to press the button declaring it so, but might not seem so good if the respondents thought about it longer.  After all, constitutional "protection" of the ABC could well make it a law unto itself; with government unable to reform it directly should internal politics drift it too far one way or the other or even cause it to waste money or churn out a plain awful product.  Technology changes could also easily see a constitutionally protected broadcaster become a national white elephant.  

Skew Poll Of The Week (Alas)

Skew-polls (commissioned polls that use biasing preambles, question orders likely to prime respondents, or other such methods to create misleading results) are so common lately that it's a challenge to keep up with them all.  Sadly the most glaring of this week's pack was by Fair Agenda, a group that promotes "fairness and equality for women".  This group commissioned a poll that asked respondents whether family violence is a bigger threat than terrorism.  See results here.  

A wide range of things kill and injure more Australians than are killed or injured by terrorism, and in many cases government responses can make a big difference to them.  So it seems like a decent point, especially given a federal government that is unusually male-centred by modern standards.  And it would be an interesting thing to see some public views on.  However, that's not what we get; what we get instead is a clearly skewed question design:

"The problem of family violence has been described as "family terrorism".  Do you think family violence is more or less of a threat than terrorism in Australia?"

The first sentence primes the respondent (the polling equivalent of "leading the witness") by encouraging them to think of an argument for considering family violence a bigger deal than terrorism.  It implies that an act of family violence is equivalent to an act of terrorism, and hence invites respondents to conclude that since family violence is obviously more common in Australia than terrorism, it must be a bigger problem.

But in fact, the line of thought the respondent is trained down is extremely contestable. A terrorist doesn't only inflict violence on their target, but by doing so attempts to intimidate the whole of society and its government with the threat of more of the same.  Domestic violence is often brutally and horribly intimidating to its direct victims, and frequently terrifies and even terrorises them, but it doesn't have anything like the same mass-broadcast element.  Nor does it have anything like the potential for attacks that cause mass casualties, or individual incidents that cause vast economic disruption.  Nor does it normally involve attacks by people on people who they have never even known.  And so on.  It might or might not be a politically or socially effective analogy, but it's a factually unsound one.

As a result of this first sentence, we don't know what Australians really think about how family violence compares as a threat to terrorism.  We only know what they say they think immediately after being placed in an artificial situation where they're fed one side of the argument and asked if they agree.

I think groups that engage in this kind of skew-polling instead of just asking it straight are doing their causes a disservice.  Had the question just been "Do you think family violence is more or less of a threat than terrorism in Australia?" without any introduction, then there would have been a credible result that might not have been much weaker than the actual outcome.  Instead what we get is a dubious slanted poll that is easily discredited by opponents.  Can lobby groups for good causes please stop being so afraid of what the public really think and just stop pulling this silly trick?

Other Issues Polling

There's not that much else to report.  Essential's government handling of issues poll found the Government was generally poorly rated.  Perceptions of the government's form had improved in most areas since January, but were down on handling of relations with other countries, and not making much progress on treatment of asylum seekers, the environment or climate change.  

Essential found respondents disapproved of protests outside abortion clinics, schools  and the houses of famous people and also inside Parliament, but thought protests should be legal outside Parliament, businesses and on the proverbial street; Liberal voters were not big fans of protests anywhere compared to other voters.

Constitutional recognition for Aboriginal and Torres Strait Islander people continues to poll well with an 85-11 response in Ipsos.

And the idea of stripping sole citizens of their citizenship over terrorism connections if they are "able" to become a citizen of another country also remains popular, with a 75-21 thumbs up in Ipsos.  Whether "able" in this context means that such a person can certainly become a citizen of another country or just that it's possible they might (at its discretion) has not been explored in these types of polls, and not much in political discussion of this issue either.

No comments:

Post a Comment

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.