Tuesday, September 23, 2014

Poll Roundup: Terror Focus Improves Abbott's Ratings

2PP Aggregate: 51.5 to Labor (-0.4 on last week, very little change in past five weeks)
Election "held now" would be very close 

This will be one of my longer fortnightly roundups, there is quite a lot to get through!  I've decided to split the mass into two halves; this is the main part and the really wonky stuff (about aggregate methods, rounding, Newspoll black magic and such) is being split off into a separate Wonk Central supplement and released to mass critical excitement later tonight.  I've made a small methods change to my aggregate, that will rarely if ever make more than 0.3 of a point difference and often make none at all. The change is documented on the methods page and all will be explained soon.

This week has seen a ramping up of anti-terrorist rhetoric and action centred on perceived threats from Islamist extremists connected to everybody's least favourite Sunni jihadists ISIL.  Massive terror raids resulted in the arrest of a man accused (and entitled at this stage to the presumption of innocence) of conspiring to order the execution of a random non-believer, plus a few other charges.  Security alert levels have been raised, military actions prepared, and a range of other measures (some of them contentious from a civil liberties viewpoint) introduced.  There's even been some rhetoric from the PM about how Australians would have to sacrifice freedom for security, although polls this week suggest voters think we're sacrificing both, and like it.

At this stage there is not a lot of public information on the basis for the threat (aside from the outcomes of the above-mentioned raids) and opponents of the Abbott regime are likely to interpret the terror talk as mostly fabricated, either to restore the PM's standing in the polls or as a front for unrelated civil liberties intrusions.  I doubt that either of these things is intentionally true, but this is Version 2 of the government that brought us Iraqi WMDs, "children overboard" and other such spurious concepts in the early 2000s, so it has a lot of work to do to convince doubters that the threat this time is not being overstated and that its response will be justified by the facts.  We'll see how that goes down the track.

In the meantime, this latest wave of terror concern has brought Tony Abbott as PM his least dire personal results since the first few months of his rule.  This is his second lift from his very bad ratings polled after the Budget, the first coming after the shooting down of MH-17.

This fortnight's polls

This week's Newspoll came in at 51:49 to Labor (down one), but off primaries of 41-34 to the Coalition, giving the Coalition its largest primary lead since late November.  The same Newspoll-watchers who were stunned when Labor gained only one 2PP point from the various moves last fortnight had no complaint about the Coalition gaining only one 2PP point from all those moves reversing and the Coalition gaining an extra point from Others.  I can only hope it's because my piece on the magic of Newspoll rounding educated them.  This time, of course, it's Coalition supporters who might be feeling short-changed by what looks like it should be at least a 50:50 to them, but more on that is coming in the Wonk Central supplement.  A general audience needs to know only this (i) it's possible (ii) it's very likely the underlying 2PP of the sample only just got over 50.5% for Labor and then rounded up to 51.  This is the Coalition's best poll from anyone since mid-April.

ReachTEL arrived late last week with a 51:49 result (unchanged on the previous month) notable mainly for the Palmer United vote falling to 5.3%, its lowest level since before the Budget.  Essential moved a point to Labor last week (53:47) and stayed there this week, showing it is now clearly off on a pro-ALP tangent following its pro-Coalition tangent a little earlier.

Finally, this week's Morgan moved half a point back to Labor to be at 53.5:46.5 by last-election preferences, equivalent to about 52:48 accounting for Morgan's house effect.  The most startling Morgan finding was a 76.5% 2PP for Labor among voters aged 18-24 (+15.5 points on the fortnight before).  Presumably the sample size for this finding was a few hundred at most, so let's wait for more before accepting that near-unanimous youth rejection of the Government's khaki footing is a thing.

All up my aggregate, with the assistance of a methods change discussed in detail in the coming follow-up, has come in from last week's recalculated end-of-week score of 51.9 for Labor to a new reading of 51.5.  While people will as usual get excited or dismayed about the Newspoll, the other polls are not showing much action and voting intention has not really moved a lot in the last five weeks.  The Opposition retains a slim lead that would be quite inconclusive in an election "held now", and that in the long term doesn't mean much at all. Here's the spiky (unsmoothed) tracking graph:

I had a reader request to look at how the exact state of polling two years out from an election predicts the election outcome, and found that on available data we can't even say that it does, and if it did the relationship would only predict, say, five percent of variation anyway.  Even just using the previous election seat outcome to predict the next one works better.  Summary data (like that the government trailed for most of its first year) might be more useful - I looked at that last time.


This week saw Tony Abbott record a Newspoll net satisfaction rating of -11 (41-52).  While this is still not very good, it is Abbott's best netsat since April (and the April Newspoll was Coalition-friendly compared to other offerings at the time.)  Abbott's satisfaction rating of 41 is his highest since late November 2013.

While Newspoll had a move mainly from undecided to satisfied, ReachTEL saw a different move with a five-point drop in the number of respondents rating Abbott Very Poor, and a 4.4 point rise in the middle-ish option, Satisfactory.  According to my ReachTEL conversion formula Abbott's ratings come out to a mere -5 netsat (also his best since November). Abbott's 30.1% "Very Poor" rating was the lowest such reading since his first ReachTEL as PM last October.

Newspoll had Abbott as better PM by just four points (41-37). the same as six weeks ago and the equal largest gap in that direction since April. This is actually an underwhelming lead for an incumbent PM in the context of a close two-party contest and nothing positive should be read into it alone.  Finally, for the record, Shorten: -5 from Newspoll (an insignificant 2-point improvement on last fortnight).  Shorten's ReachTEL was hardly changed from last month (converting to about +1) but the number picking the nearest thing to a "meh!" option, Satisfactory, was at its highest level since February.

Issue Polls: Non Terror-Related

There's quite a lot of issue polling out so I'm going to start with one that's a little bit different.  Following one of Jacqui Lambie's few remotely sane suggestions of the past fortnight, Fairfax Media commissioned ReachTEL to survey attitudes to the creation of Indigenous seats in parliament.  (This is something that sounds good in theory but would require a referendum in practice, and there are arguments that it is potentially perversely disempowering, since often left-leaning indigenous votes could get coralled into a small number of safe seats, reducing their impact on outcomes elsewhere.)

I make the result for Independent/Others voters something like 26% Yes, 46% No.  Generally, what we see here is that the more left-wing a voter is the more likely they are to agree with Sen Lambie's proposal.  That's not surprising but what is remarkable is the massive gender gap - females 14 points more likely to support, 20 points less likely to oppose.  Some of this is because female voters are slightly more left-leaning. Possibly the issue is seen as a proxy for affirmative action more generally but I doubt even that totally explains it.  Suggestions welcome.

Essential respondents oppose buying submarines from Japan even if it is cheaper (51:28) with Coalition voters evenly split and ALP voters (19:60) more against than Greens supporters (28:57).

Essential attribute polling found perceptions of the ALP almost unchanged in six months, save that voters thought the party marginally less divided and slightly less sure what it stands for.  The Coalition was oddly seen as more divided (probably just a rise of blanket naysaying rather than any actual basis in evidence for this) and its scores worsened on every attribute, most notably "keeps its promises" (down 9).

Essential also found perceptions of the government's performance were negative on every issue except "Relations with other countries" (+15) and had worsened since February on every issue except that one and supporting Australian businesses.  The biggest decreases were for typically left-wing issues: education, health, welfare and climate change.

These findings should be seen in context of Essential's very Labor-friendly polling lately and hence treated with some caution.  But ReachTEL also found voters tended to rate the government's performance in the first year a bit dimly (I convert it to a net -10 rating), and that voters don't think the government has kept its promises (52:27 with the usual partisan split.)

Essential found voters disapprove of the shelving of plans to increase compulsory superannuation levels (29:49) with only Coalition supporters approving. It seems difficult to sell the message that increased compulsory super is a potential impediment to hiring or wage rises, with almost nobody believing the change would lead to greater wage rises (10:65).

Much fun can be had with the Guardian's interactive of the McNair Ingenuity poll of ministerial ratings, which shows that Coalition and Labor supporters align in rating Julie Bishop the government's best performer.  Bishop's score of 57 on this method is a rather good one, as the method used tends to compact scores towards 50%.  It's probably worth about a +15 net rating.  That doesn't say much for those rated at the tail by supporters of both parties (Brandis, Andrews, Abetz, Pyne and at the bottom of both heaps, Hockey with a dreadful rating of 32.2).  Indeed none of the ratings bar Bishop's are anything special.

Turnbull, Johnston, Joyce and Robb are near the top for both parties.  Morrison rates very highly among Coalition supporters but well down for Labor, and Abbott quite highly for his own party but near the bottom of the pile for his opponents.

Issue Polls: The Khaki Corner

Essential's respondents approve (52:34) of "sending 600 military personnel plus aircraft to Iraq for action against the Islamic State (IS)", with Coalition supporters in favour, Greens against and Labor split.  Coalition supporters agree that the main aim of the game is to fight terrorism, Greens supporters think it is to support the USA, Labor supporters are again split. Most voters (51:15) think Australia will be less safe as a result of doing so; this time all parties agree.

Ipsos finds an even more extreme split on the last question (58% believing less safe vs 3% more safe). Ipsos found an overall 47:33 response in favour of deployment, and a 43:41 rating for the government's perfomance on foreign affairs overall.

ReachTEL separated the questions, with voters supporting sending planes to Iraq (49.1:29.3) and opposing ground troops (32.8:44.1)

Essential supporters weakly oppose (36:42) Australian military aid to Ukraine, with only Coalition voters (54:32) in favour.

Morgan SMS Issue Polls

I've quarantined these ones in special detention.  Morgan have been lately issuing snap SMS issue polls on issues like sending troops to Iraq (54% support 46% oppose, with a very big gender gap) and the East-West link tunnel in Melbourne (63.5:36.5 in support of construction).  Oh, and burqa-banning (55.5% reported to support banning to 44.5% oppose,  with a very strong corellation between supporting banning and age.)   Well, I'm not so sure about this polling method so far.  There is apparently no undecided vote so I can only assume that the voter is sent a yes/no question by SMS and the undecideds or those whose opinion is too weak to be bothered responding don't text back and are ignored.  It might not sound so different to ReachTEL's "answer forcing", but it is if there's only one question, because ReachTEL voters have already invested effort in answering the phone and the first question, which becomes wasted effort if they refuse to answer later questions.

My concern is that this method is a bit too opt-in and prone to motivated response, though I expect answers are at least demographically targeted and/or weighted to eliminate the main problems with opt-in surveys (even from a representative base).  Anyway I am treating results from this method with caution until we know more.


I hereby post a completely gratuitous link to Michelle Grattan's Dean Jaensch Lecture on the first years of Australian PMs since Whitlam.  Yours truly is acknowledged there for work on historic Morgan polling, as also documented in the Advanced Dungeons & Dragons module "An Adventure To Iron Mountain".

(Sorry, that was an injoke for U of Tas library users.)


  1. This comment has been removed by a blog administrator.

  2. This comment has been removed by the author.