Poll Roundup: Return Of The February Wobbles

2PP Aggregate: 53.0 to Coalition (-0.3 this week, lowest since early November)
Coalition would comfortably win election "held now", but could lose some seats

Last week, the Coalition government under Prime Minister Malcolm Turnbull was forced to reshuffle its ministry.  This became necessary following the loss of Human Services minister Stuart Robert to concerns regarding an "unofficial" trip to China, the delayed resignation from the ministry of Mal Brough (under police investigation) and the announced retirements at the next election of Warren Truss and Andrew Robb.  While hardly the worst week a sitting government has endured, it's not something they'd want to repeat. The loss of three ministers to scandals since Mr Turnbull took over hardly helps create an image of post-coup stability, let alone the mirage of "good government".  Admittedly, none of them were major figures.  A second problem for the government has been a perception of planlessness in the conversation it started about tax reform.

The four polls in the last week have all suggested the government has come down a little from the cloud that it started the year on.  This week's Ipsos has only 52:48 to the Coalition, down from 56:44 in November.  ReachTEL has 54:46, down from 55:45 a few weeks ago; Morgan is down from 54 to 52.5 by last-election preferences, and Essential is stable at 51 but off marginally worse primaries.



The Ipsos figure is by "overall preference flow" (treating Greens and Others as a single block).  At the last election, Green votes were 41% of all non-major-party votes but in this Ipsos they are 60% of the non-major vote.  On this basis the last-election preference flow of Ipsos' figures with Greens and Others subdivided would be about a point lower than their calculation, so there's a case for treating the Ipsos as 51:49 (if you really believe 15% for the Greens, at least).  I am going to explore altering my hybrid primary/2PP aggregation methods to adjust for Ipsos' use of "overall preference flow" but for the time being I've counted the Ipsos as 51.6 and the ReachTEL as 54.4 after considering the primaries.  ReachTEL is now recording more Coalition-friendly readings than the other polls, but not yet by enough consistently enough to cause me to do anything about it.

I changed the estimate of house effects of both Morgan and Essential in my model last week and as a result the Morgan was aggregated at 51.3 (would have been 50.8 before the change) and the Essential at 52.8 (would have been 53.3).  All up the Coalition has fallen below 53 for the first time in 14 weeks:

(Other aggregators post-Ipsos: Phantom Trend 53.5 Mark the Ballot 53.4.)

This is not exactly "panic stations" polling, but it's interesting to observe a historical trend - recent governments have often tanked in the polls around this time in an election year.  Julia Gillard had the Week From Hell in 2013 dropping over two 2PP points on the spot (premature election announcement, Nova Peris "captain's pick" brawl, retirements, disappearing mining tax revenue, Swan figures gaffe etc).  However, Labor were also tanking at this time in 2010, under pressure from new opposition leader Abbott as Kevin Rudd's lack of carbon courage because clear to the electorate.  The Howard government was also getting kicked around now in 2004 and 2007 (in both cases with new Opposition Leaders in honeymoon mode) and was about to experience one of the most infamous slumps in polling history at this time in 2001 (though in that case the slump came in early March).  Indeed to find the last times something like this didn't occur, we have to go back to the cycle of elections from 1990 through to 1996 in which the election was much earlier in the year.

This history, of course, tells us nothing about whether the current slide is temporary or permanent.  All the above slides have had their own causes, but none have been in the wake of a change of Prime Minister.

One other thing notable in current polling has been the very high Green vote in both Morgan (16) and Ipsos (15) compared to Essential (11), ReachTEL (10.1) and Newspoll (11).  Overpolling of the Green vote is a pet subject on here and I just don't believe the mid-teens figures at all.  If anything, I wonder if the problem of overpolling the Green vote may be getting worse for some pollsters - as respondents get more difficult to find, Green voters (who tend to be more interested in politics than most) are probably much easier to oversample.

Leaderships

The general picture stays the same here: Malcolm Turnbull is very popular (though not as much as late last year), Bill Shorten is unpopular, and Turnbull has the kind of thumping lead as preferred PM to be expected based on the 2PP, the leaders' personal ratings, and the way such figures skew to the incumbent.

Specifically:

* Essential has Turnbull on +24 (51-27), down two net points in a month.  Shorten is on -21 (27-48), down one point.  Turnbull is preferred as PM 52:15 compared to 51:18 a month ago.

* Ipsos, which last polled in November, has Turnbull on a cruisy +38 (62-24), but this is still down fifteen points on his November level.  Shorten is up three to -25 (30-55).  Turnbull is preferred PM 64:19 compared with 69:18 late last year.

* ReachTEL's poll last month was problematic, and the issues with the leadership figures in that poll have not been resolved or explained, so I am ignoring the extreme January results and using November as a comparison point.  The current poll does not have the same issues at all so I suspect a once-off glitch.   The current poll has Turnbull on a net rating of 15.3 (down 18.1 since November, similar to his decline in Newspoll), Shorten on -31.7 (compared to -26.9 and a new worst from this pollster if we ignore the suspect January result), and Turnbull preferred by 74.9% in a forced-answer choice (intriguingly, up slightly on November and the widest gap so far with January ignored).

I notice that the percentage disapproving of Turnbull's performance is more or less the same in Ipsos and ReachTEL but the former has a higher percentage approving, consistent with soft-approvers using ReachTEL's "Satisfactory" option.  This is probably also happening with Shorten.

Shorten copped another ReachTEL brickbat with a preferred-ALP leader poll showing him on 22.3% trailing Tanya Plibersek (30.5%) and Anthony Albanese (24.7), with Chris Bowen (13.2) and surprise novelty selection Tony Burke (9.3%) rounding out the field limited to five choices.  (Whoever wanted Burke included as a leadership contender has a very short memory of his travel expenses form, which came out just after the Bronwyn Bishop expenses scandal).  The similar Not-A-Poll on the sidebar here has had Albanese and Shorten trading the lead a few times in recent weeks with even Mark Butler having a turn in the lead at one stage to my surprise (though it's all for fun and not entirely impossible some very bored person is gaming it).

ReachTEL also had a poll regarding new Nationals leader Barnaby Joyce (hmmm, am I really writing that?) as the new Deputy PM, with respondents split more or less one-third each on whether he'd be good, bad or indifferent.

Other Polling

Essential's polling on the current tax debate last week shows that it hasn't been changing anyone's mind, possibly because voters don't even understand the reasons for it.  Support for increasing the GST if taxes are cut is as partisan as it always was.  Limited reporting of a Galaxy poll finds its findings on support for ending superannuation concessions (57%), ending negative gearing (39%) and increasing the GST (24%) (the latter presumably without the wording mentioning any offset) to all be virtually identical to Essential's.  Newspoll found 62:27 support for increased taxation of super contributions for high-income earners.

Both Essential (36-31) and Newspoll (51-37) found modest leads for Australia becoming a republic, but nothing likely to stand up to a referendum environment.  Predictably, support increases if premised on the Queen no longer being on the throne (45-29 and 55-34).  Essential also found strong opposition to changing the flag, the national anthem or the date of Australia Day, though Greens voters sympathise with the first and third of these concerns.

Essential also found that trust in three major Murdoch papers has increased, with the Courier-Mail and Fairfax majors not enjoying such fortune.

I think that's all for this week, another week in which I have gone early with the roundup for work reasons and will update with polls through this week as they arrive and as time allows.

Tuesday update: Essential came out at 52:48, which may have been good news for the Coalition given its recent skew to Labor, or may be a sign that excursion is ending.  Anyway that moved the Coalition back up to 53.