Poll Roundup: Turnbull Popular, Size Of Lead Less Clear

2PP Aggregate: 51.3 to Coalition (-1 in a week, +4.9 since final under Abbott)
Coalition would easily win election "held now"
(Aggregate updated from 51.5 to 51.3 following Essential)

Four weeks since Malcolm Turnbull rolled Tony Abbott, it is clear that the change is being well received.  Polls continue to show strong approval for the change, strong personal ratings for Turnbull and massive leads for Turnbull over Shorten as preferred or better Prime Minister.  What is less clear is how voting intention is travelling, and the uncertainty is coming mainly from the strange behaviour of a single pollster (Morgan), together with a shortage of data from others.

Polls over last three weeks

Since the previous roundup (Turnbull Shift Puts Coalition Back In Front) we've seen two weekly Essentials, one Morgan and one Newspoll.  A third Essential will be out tomorrow but because of work commitments I've decided to release the roundup now and update it for Essential sometime tomorrow night.

The two Essentials both had headline rates of 52:48, but for the first one a weekly 2PP of 53.5:46.5 was released, suggesting that last week's sample was probably around 50:50.  The Newspoll this week is at 50:50.  Last week's fortnightly Morgan, however, came out with an off-the-scale result of 55:45 to Coalition by last-election preferences (56:44 respondent-allocated).



The Newspoll has caused some nervousness among Coalition supporters because it does not show a 2PP lead.  It is worth noting here that Newspoll has been slightly Labor-leaning since the Galaxy takeover, and also that 50:50 would very likely be a win in an actual election.  Nonetheless there is a line of thinking that if some of this is still a honeymoon bounce, then the Coalition could find itself back in trouble in the future.

With adjustment for the primary votes, but without applying any house effects, I have the Essentials at 51.6 and 51.7 to Coalition, the Newspoll at 50.1 and the Morgan at 55.2.

Aggregation

The super-strong Morgan result for the Coalition (combined with the slightly weaker dose of the same the previous fortnight) is odd because through this term Morgan has leaned to Labor, by an average of slightly over a point once primaries are taken into account.  Now, following a change of Prime Minister, it has produced two very strong results for the Coalition.  Some might say that this is just two rogues in a row and that Morgan is unusually prone to big outliers.  But the latter hasn't actually been true in this term once house effects are taken into account.  Also, while the "rogue" range starts at just under two standard deviations from the mean (a 1 in 40 chance in either direction), Morgan's last two results come out at over three SDs astray, making the probability of these two results appearing by chance extremely low.  (One in hundreds of thousands, or even less, depending on assumptions.)

I've assumed on this basis that the house effect of Morgan's polling has been changed by the shift to Turnbull, but it's hard to know how much it has changed by from just two results.  For the time being I've downweighted Morgan in my aggregate and removed the previous house effect to Labor.  I've also back-applied these changes, meaning that the last three end-of-week readings under Turnbull change from 51.3 to 51.1, from 51.8 to 51.4 and from 52.6 to 52.3.  This week's is currently 51.5, to be updated after Essential.

The potential problem with Morgan is that both its polling methods could produce distorted results in the current situation.  Face-to-face polling runs the risk of the respondent wanting to agree with whatever they think the interviewer thinks or wants to hear.  SMS polling might be prone to "motivated response" - the tendency of those with the most enthused or strongest views to take part in the survey.  I have less trust in either of these methods in a novel situation than in others.

It may be that Morgan now has a house effect in the Coalition's favour, in which case my estimate for how the Coalition is travelling under Turnbull may be a bit high.  If I removed Morgan from my aggregate altogether, the current reading would be 50.6, but it's always possible that the other pollsters are collectively partly at fault for the differences between them and Morgan.   Either way, the Coalition would win an election held now as they probably only need something in the high 48s. Here's the smoothed tracking graph, noting again that no smoothing is applied across the Abbott and Turnbull results:

The weekly readings are pretty volatile at the moment because of the Morgan issue and because there have been few polls lately, so I wouldn't pay any attention yet to the apparent peak last week.  What we have so far is consistent with a boost of about five points (but it might really be four or six), and it will take a while to be clear about when it might peak and what it might do after that.

William Bowe's analysis (largely paywalled) shows that the boost in Coalition primary support is coming among median-age and especially older demographics, and to a large degree at the expense of the Greens.  It is no surprise if the higher-income end of the Greens surge under Abbott was always soft.

Other current aggregate readings are: Phantom Trend and Poll Bludger 51.2 to Coalition, Mark the Ballot 51.7.

Leaderships

This week's Newspoll produced some more notable results.  As voters start to form impressions of Malcolm Turnbull as Prime Minister, those coming off the fence are responding mostly favourably, giving Turnbull a netsat of +25 (50-25).  This is the highest netsat polled by a Prime Minister since Kevin Rudd polled +28 in early December 2009.  Turnbull's satisfaction rating of 50 is the highest since Julia Gillard polled exactly that in February 2011 (Rudd last polled above 50 in March 2010).  Turnbull also stretched his Newspoll lead over Bill Shorten from 55-21 to 57-19.  This is now the largest lead for a sitting PM since Rudd led Turnbull 65-14 in November 2009.

Bearing in mind that the polling methods of Newspoll have changed following the transfer to Galaxy, the most notable feature of the Newspoll is that Turnbull is so popular yet his party is only level on 2PP.  Here's a graph of Opposition 2PPs (in some cases Peter Brent's old estimates) for given PM netsat levels with the current result highlighted as a red dot on the edge of a sea of blue:

It's extreme, but it isn't a record yet.  In late 2002 following the Bali bombing, there was an initial boost in both the Coalition's 2PP and Prime Minister Howard's personal ratings.  The latter boost took longer to decline than the further, and produced one poll reading with Howard at +27 but the Coalition at 50-50.  For whatever reason a pattern of high PM personal ratings but overall slight Coalition leads then cropped up a few more times during the Howard-vs-Crean period.

Quite likely, this disparity won't last. To the extent that Turnbull repairs the government's image, the 2PP should rise, and to the extent he fails his own ratings should fall.  But it's possible that the damage done by the Abbott era will continue to distort the usual link between the PM's popularity and his party's all the way to the next election.

Shorten's Newspoll rating was again pretty awful at -25 (28-53).  Over at Essential, Turnbull is at 47-17, Shorten at 30-42 (a nine point net improvement from September) and Turnbull leads Shorten 48-19.

In other polling of passing interest, Essential found that voters wanted an election later next year, and found that party trust to handle issues is really not much changed by the switch to Turnbull, with the sole exception of the issue of leadership.

No-One Much Listening To Abbott Rump

The last few weeks saw several attempts by Tony Abbott and forces loyal to him to re-establish his legacy.  The polling response to these pro-Abbott agitations has been lukewarm to say the least. Newspoll found massive approval of the removal of Tony Abbott as PM (62% to 27%, with Coalition voters approving 56-36).  Labor supporters approved 71-22, Greens 82-11 and I infer Others approved only mildly, 32-25.  (The likelihood here is that a few percent of these Others really have no idea who they'd vote for.) Essential respondents were split 41-41 on whether Abbott should even remain in parliament at all.

Andrew Bolt has taken the popularity of the change among Greens voters as a danger sign since Greens voters are "most unlikely to vote Liberal".  What Bolt ignores here is that a Green voter who switches from preferencing Labor to preferencing Liberal is as much a gain as a voter who switches from Labor to Liberal outright.  Even a small shift in Green preferencing patterns can be worth a few seats.  Bolt also ignores the evidence that quite a few voters have already moved between Green and Liberal as a result of the change.  (For some not especially right-wing voters there is a basically classist aversion to voting 1 ALP.)

To the extent that they touched on polling, Abbott's claims were: firstly, that David Cameron had won despite persistently trailing in polls;  secondly, that Abbott had recorded good election results while being unpopular before and could have done so again, and thirdly, that polling for the Canning by-election showed it would have been won easily, undermining the case for his removal.

The case being made by Abbott thus far is pretty weak, but there is just enough in it to reinforce the views of those who already want to believe.  For starters, the polls Cameron trailed in were actually wrong, and that there were many UK-specific issues with them that mostly don't apply here.

Abbott was actually not that unpopular in the leadup to the 2010 election (polling net minus single figures through the campaign, and no worse than -15 in the leadup) and even by the time of the 2013 election, he had recovered to -6.  More importantly, however, Opposition Leader ratings have only a very weak, if any, relationship to 2PP voting intention, while PM ratings have a very strong one.

The latter also holds up at elections: since leader ratings were first polled in the late 1960s, every PM bar one (Howard 2007) who has gone to an election with personal ratings of net -11 or better has won, and every PM bar one (Keating 1993) who has gone to an election at worse than that has lost. So there is no reason to think that Abbott would have won with a bad rating.  He might have improved his personal standings as an election approached, but no case to that effect was made.

As for Canning, the 57:43 internal poll (with Abbott still PM) first touted by Andrew Bolt apparently did exist, as part of a series that bottomed out at about 53.  But it's not remotely likely that voting intention for that by-election moved around that much during the campaign, so if the final Liberal internal under Abbott was really 57:43, then it was most likely an outlier and not a reliable measure of the likely result (as if a single internal seat poll - three levels of uncertainty piled on each other there - would ever be so anyway.)

My probably final view on Canning is that the Liberals would have still won it under Abbott, with at least a 53:47 margin.  Had it been 55:45, this would have been a good result for Abbott and raised some eyebrows about poll reliability, and perhaps bought him a bit more time, but it wouldn't have been a game-changer.

As for Abbott, I don't think he was doomed to certain defeat had he been retained.  But there was ample evidence that his leadership greatly increased the risk.   It will be interesting to see where this pro-Abbott agitation goes over the next year, but for the moment it is having little if any impact.

Not-A-Poll

There is a Not-A-Poll on the sidebar where people can vote for which pollsters they think are most reliable.  Each respondent can choose up to four.  Unfortunately the poll software doesn't say how many people have voted, so while Newspoll is currently on 24% of votes, it's likely that it has been included in something like 80% of votes cast.  At present Newspoll 23 leads Galaxy 21, ReachTEL 19, Essential 13, Ipsos 10 and Morgan 4.  ("None of them" has 3).  This poll is very stackable so don't be too surprised if one poll someday gains 50 votes after all its employees vote for it (so far this hasn't happened.)

Essential Update Oct 13

Essential has come out with a two-week rolling average of 51:49 to Coalition, which I aggregated at 50.9 after considering the primaries.  What this actually means is not that this week's sample was worse for the government than last week's (it was probably actually better) but that this week's sample was not as strong as the one from two weeks ago.  Anyway this knocks another 0.2 points off this week's reading, or 0.1 of a point (to 50.5) for the Morgan-free version, which I will continue keeping an eye on for a while.

Essential finds fairly strong approval for the Trans-Pacific Partnership, concentrated among major party voters with other voters mildly opposed.  Essential also returned a finding that voters don't like contracting out government services, as well as typical Essential findings that voters think the terrorism threat has increased in recent years (75% think this) but tend to expect air strikes against Islamic State to make the problem worse.