Monday, August 9, 2021

Poll Roundup: Lockdowns And Rollout Problems End Morrison's Net Rating Streak

It's been a while since my last federal polling roundup and in that time the government's troubled vaccine rollout and the challenges polled by the Delta strain of COVID-19 seem to have put paid to government optimism about rushing off to an early election this spring.  Indeed, recent weeks have seen the government record its worst 2PP polling of the term so far, though by historic standards it is not yet in severe poll-based trouble.  

Voting Intention

Three weeks ago, four polls came out in quick succession:

* A Newspoll Coalition 39 Labor 39 Greens 10 One Nation 3 Others 9.  The published 2PP was 53-47 to Labor but Labor probably got lucky on the rounding here.  The average last-election 2PP for these primaries would be 52.5-47.5, and indeed when exactly the same primaries came out in early March, the 2PP was 52-48.  Given that this one rounded to 53, it must have been at least 52.5 prior to rounding.

* Essential Coalition 37 Labor 36 Greens 10 One Nation 4 Others 6.  After rescaling to remove undecided and get sum to 100, Coalition 39.8 Labor 38.7 Greens 10.8 One Nation 4.3 Others 6.5.  The published 2PP is 47-45 (=51-49) to Labor under Essential's 2PP Plus method using (mostly) respondent preferences, but by last-election preferences I get 52.1 to Labor.  

* Resolve Political Monitor Coalition 38 Labor 35 Greens 12 One Nation 4 "Independents" 7 Others 5.  Resolve does not publish a 2PP.  By last-election preferences these numbers come to about 52.0 to Labor if taken as published, but Resolve overestimates independents.

* Roy Morgan Coalition 39 Labor 37 Greens 11.5 One Nation 3 Others 9.5.  Morgan's published 2PP was 52.5 (respondent preferences); I get 52.0 (last election).  

In the last week two fresher polls have arrived:

* Another Newspoll with a 53-47 2PP to Labor, but this time with the Greens on 11 and Others on 8, which all else being equal suggests the 2PP this time was probably close to 53 prior to rounding.

* A Morgan with a 53.5% 2PP to Labor off primaries of Coalition 37 Labor 37 Green 12.5 One Nation 3 others 10.5.  I get 53.3 (last election).  This is the highest 2PP poll for Labor of the term except for some bushfire-era Morgans for which only 2PPs were back-released. 

I am not currently running a polling aggregate (still thinking about what on earth to do with house effects in light of what we saw in 2019) but if I was running one without any house effect assumptions, it would probably put Labor on about 52.7.  If one believes these polls, Labor would win an election "held now" by a margin of something like 81-64-6 (a very likely but probably modest majority), but after 2019 it is possible that this is wrong and the government might still be somewhere around level-pegging - for now.  

I have had a look at house effects for the current polls relative to each other and I find that the following poll 2PPs have been more or less equivalent this year after adjusting for the time they were taken: Newspoll, Essential (respondent preferences after removing undecided), Morgan (respondent preferences), Resolve (derived 2PP treating independent as stated).  So I see no real evidence that any poll is producing more favourable results for either side than another in a 2PP sense, but there is a tendency for Newspoll to have the major party primaries higher than Morgan and Resolve.  Based on the history of polls that have high minor party and independent primaries, I am more inclined to believe Newspoll.  

Is Resolve irresolvable?

It is difficult to reliably convert Resolve's primary figures to a 2PP estimate because the poll is overestimating independents (finding the "independent" vote far higher than at the last election federally and in both NSW and Victorian state polling).  This could result from any or all of:

* Independent being on the simulated ballot everywhere, when actually independents only run in about half of all seats.  (This is unlikely to explain all of it, as the independent vote is concentrated - almost half of it in 2019 occurred in just nine divisions).  

* Voters saying "independent" when they actually mean a micro-party.

* Voters parking their vote with "independent" when they are actually going to vote for a major party, probably Labor.

* Voters who will vote for some kinds of independents but not all.

* The poll either over-sampling or over-weighting voters who intend to vote independent. 

Some of these factors suggest that treating "independent" as independents could be harsh on the Coalition (since micro-party voters preferenced the Coalition 59-41 in 2019) but it's also possible these would-be "independent" voters are lefties, perhaps even more so than actual independent voters who preferenced Labor 60-40.  I've switched to treating these "independent" preferences as stated for now - treating them as generic "others" is also an option and usually results in a 2PP that is close to a point better for the Coalition.  

Trying to figure out what is going on here is not helped that much by the publication of statements like "The Resolve Political Monitor provides a suite of minor party and independent options in our primary vote measure, just like the real ballot paper."  In fact, whether the "real ballot paper" contains a "suite" of such options or not at any given election depends on where you live, so in the absence of sufficient detail, I can only doubt whether that statement is true.   I also note that Essential's methods statement says they have One Nation on the readout everywhere, which will overestimate One Nation's vote if the party don't actually run everywhere, and did so at the last election.  

I have frequently mentioned the problems with the lack of a 2PP estimate in Resolve's polling.  More evidence that the quest for finding something "deeper" by abandoning 2PP is on the rocks comes from the SMH letters page, in which a number of readers, provided with not even a hint that Labor would probably win an election on Resolve's numbers, did their own horserace and despaired that Scott Morrison is still winning because of primary votes and, even worse, "Preferred PM". 

Leaderships and COVID handling

Scott Morrison's personal ratings have been dropping sharply in recent Newspolls and this weekend saw him finally record a negative net satisfaction rating, within the window predicted by almost half of the few voters in the sidebar on this site (not sure what those who picked "Never" were thinking!)  

This ends a streak of 21 consecutive positive Newspoll netsats for the PM.  This is the fourth longest ever in terms of number of polls, and the third longest ever in terms of time span.  However he is a long way behind the leaders.  The top five streaks, as noted with many other such things on my Federal Newspoll records page, are:

102 John Howard (Coalition), Aug 2001-Oct 2005

53 Kevin Rudd (Labor), Jan 2008-Apr 2010

28 John Howard (Coalition), March 1996-May 1997

21 Scott Morrison (Coalition), April 2020-Aug 2020

16 Bob Hawke (Labor), April 1987-December 1987

As Morrison's ratings have slipped there has been plenty of talk about this meaning that the voters have "found him out", especially over a habit of avoiding responsibility when things go wrong.  There have been plenty of unflattering comparisons to Howard as if Howard would never have lost the plot like that.  But this is what the historic record of Howard's and Morrison's ratings shows (Howard in blue, Morrison orange):

Morrison is now in his second significant crash (the first was the 2019-20 bushfires).  No-one knows yet where the bottom of his current decline is and it may be that it becomes a very large one.  However at this stage Howard had plenty of declines that were larger and/or steeper, with significant falls in support in  1996, 1997, 1998, 1999/2000, 2001, 2002, 2002/3 and 2005.  Indeed it was only after Howard's ratings stopped going up and down so often that he finally lost!

Nonetheless it has taken Morrison only 16 months to go back under the water after the massive bounce at the start of the pandemic.  Howard's positive streak started with the MV Tampa incident and then, amid a steady feed of khaki politics and strong economic results, he managed to keep it going for over four years.  Had anyone predicted this in March 2001 when he was languishing on net -36 no-one would have believed it, and this should be a warning about how unpredictable the public view of leaders is.

Resolve Political Monitor has also had Morrison's ratings falling (indeed it slipped into net negatives there from a lower base three weeks ago) though Essential still had him on net +10 this week (with a curious fall for Anthony Albanese to a new personal low from said pollster instead, which Newspoll hasn't replicated).  

While it is not the only contributing factor, I see the current lockdowns/vaccine rollout nexus as the main cause of Scott Morrison's drop in popularity.  Whether it is actually correct or not, the argument that we would not now be seeing an extended lockdown in NSW and a spate of short lockdowns in some other states if the vaccine rollout had been faster is an easy one for people to agree with.  Moreover, it is even easier to argue that there will now be lockdowns over a longer period of time later this year than would have been the case had Australia matched other countries.  In these circumstances the goodwill accorded the government for Australia's relative success in keeping the pandemic in check seems to be rapidly waning.

This can be seen with Newspoll polling specific to handling of COVID, where net approval of the PM's performance has crashed from nearly universal (85-14) last April to net negative (48-49) now and perceptions of the federal government's handling of the vaccine rollout are now deeply negative (38-59), especially among women (34-62).  

Morrison's lead on the skewed Better Prime Minister metric also continues to fall - he is now down to a lead of 13 points (49-36) over Anthony Albanese, compared to a 35-point lead in February.  Morrison's lead is now slightly below the historic average lead to incumbents of 14.6 points, and also below what would be expected if the parties were level-pegging.  However, that lead is still outperforming what would be expected based on the 2PP and given that Morrison himself is no longer personally popular.  Whether this says something about perceptions of Anthony Albanese (such that an unusually large number of pro-Labor voters may think he is satisfactory as an Opposition Leader but not PM material), or whether it is still something about the pandemic and/or Morrison I'm not sure of yet.  However the fact that Morrison's net personal rating is also higher than expected for the 2PP points away from it being about Albanese.

How much trouble is the government in?

I covered this briefly in last week's article about the 2022 pendulum, but it is worth mentioning again.  The government's current polling deficit does not mean it will necessarily lose or is particularly likely to lose in 2022.  Governments that are going to lose will typically fall much further behind in polling than this at some time.  They habitually escape from anything above 45-46% 2PPs and sometimes from much worse than that (see here).  The concern would be that negative perceptions of the government will harden through coming months as lockdowns drag on (especially in NSW) and that the damage might be too much to recover in the time remaining before polling day, even if the causal issues go away and COVID comes under a measure of control early next year.  Also, a number of the past dramatic recoveries by governments (1993, 2016 and 2019 especially) have been from polling that was dire in either the lead up to, or the immediate aftermath of, leadership change on the government side.  While it's tempting to conclude that the government is going to face much worse trouble in times ahead, polling from overseas (especially UK) shows that it is risky to assume that COVID situations getting worse will damage governments.  Voters sometimes respond by rallying behind the government as soon as there is any kind of good news whatsoever, or sometimes even when there isn't.


And now a quick note on where the potential transparency revolution following the formation of the Australian Polling Council is up to.  In the case of YouGov/Newspoll, there is an excellent page with public disclosure statements for some polls you may not have even known existed - this firm is doing a terrific job of complying with the APC requirements.  In the case of Essential, there is a page up with a link to a more detailed statement, though what I have seen so far only goes into the mechanics of the regular voting intention and leadership questions and not the issue questions.  (I have often wanted more information on Essential's issue questions methods - in particular re the order that answers are offered and whether or not it is randomised.)  To my knowledge, only these two pollsters have had disclosure requirements with their national polling so far.   

Unfortunately, I'm still unable after a few weeks to find a full long methodology statement anywhere for a uComms that included Tasmanian Senate polling and was clearly commissioned by the sponsor with the deliberate intention to put results in the public domain, which the sponsor then did. (An incomplete LMS was published by an independent website, which in any case doesn't count.) 

It's this kind of poll I think it's most important to see much, much more transparency progress on, because one-off commissioned polls are usually the hardest to find sufficient details of, the worst designed (skewing preambles especially) and the worst reported.  Nearly always these dubious lobby group polls get an easy ride from mainstream media reporting on them. Journalists, either grateful for an easy story or simply having no time, seem less and less inclined to get independent comment on the polls, but so often the claims made by the sponsors of such polls are biased and unsound and are yet passed off in mass media articles as actual facts about public opinion.  

As concerns Resolve, some insight into Resolve's reluctance to join the Council came from a forensic (and excellent) article about Resolve's recent polling by Murray Goot.  At the end, Goot reports that the pollster is reluctant to divulge "trade secrets".  Firstly, following what happened in 2019 there should be an expectation from the public and the media alike that all polls - involved in that failure or not - will tell us a lot more about what they are doing if they want us to take their results and reports all that seriously.  Secondly, the Polling Council standards are a small amount of work for pollsters to comply with, but they are simply not onerous in terms of requiring pollsters to declare "trade secrets".  Pollsters are not required to divulge their 2PP calculation formulae, their weighting formulae or even their raw sample data, let alone the pages of crosstabs routinely seen in UK polling reports.

No comments:

Post a Comment