Monday, January 13, 2020

Newspoll Roasts Morrison / 2019 Polling Year In Review

Newspoll has come out of hiding early this year, and that warrants a quick post about the unusual nature and results of this week's early Newspoll, to which I am also attaching a belated annual roundup for 2019.

In the past, the history of Newspoll has tended to show that national security related incidents have big impacts on polling, but natural disaster incidents generally don't.  (An exception was at state level, where Anna Bligh's doomed Queensland Premiership received a large but temporary bounce from her perceived good handling of the 2010-11 floods disaster.) However, this natural disaster is somewhat different, both because of the scale of its many impacts and the extent to which lines of criticism of the federal government have immediately opened up.  Prime Minister Morrison has been criticised for taking a holiday during the crisis, for insisting on shaking the hands of bushfire victims who didn't want their hands shaken, over the level of federal preparation for the crisis, and over the government's climate policies and degree of recognition of realities of climate change.

Some of these criticisms, especially the last, are coming mainly from people who did not support this government anyway, and so it was hard to say what the impact on the government's standing might be until we had some numbers on it.  Even then, we should treat these numbers with some caution, not only because of the relative failure of polling in last year's election, but also because it is unusual to see polling at this time of year.  In fact, the polling schedule (8-11 Jan) was the earliest out-of-field date in Newspoll history by two days.  Furthermore, it has been unusual in recent years to get Newspolls in January in non-election years at all.  So it does look like the interest value of the bushfire situation could have resulted in Newspoll going back into the field earlier than normal.

While the three-point drop in the government's 2PP polling from 52-48 to 49-51 (Labor's first lead since the election, not that wall-to-wall leads before it did them any good) is of some interest, the real headline-grabbers here are the leadership ratings.  Firstly, Scott Morrison is down 19 points on net satisfaction from -3 to -22 (37-59).  This represents:

* The equal sixth highest poll-to-poll drop in a PM's netsat.  The two highest are 24 points by John Howard in a rogue poll in February 2002 and 23 points by Julia Gillard after she reneged on her "no carbon tax" pledge in March 2011.  Hawke, Keating and Rudd each recorded one loss of 20 points with Keating's coming after the 1993 horror budget.

* The equal worst dissatisfaction score that was recorded by Morrison's predecessor Malcolm Turnbull.  However, this is not comparable as the new Newspoll methods have reduced the undecided rate on Prime Ministerial satisfaction - Turnbull's worst netsat was -30.  (Several PMs prior to Turnbull had recorded worse dissatisfaction scores than 59).

Anthony Albanese is up ten points to +9 (46-37).  This is not only his best result so far but also represents the highest satisfaction rating for an Opposition Leader (46) since Tony Abbott recorded 48% immediately after very nearly winning in 2010.

The Better Prime Minister indicator is normally skewed to incumbents, especially if their parties are not too far behind, but in this case Anthony Albanese leads Scott Morrison by 43-39.  This is an 18-point turnaround on the margin since the last poll.  This represents:

* Albanese's first lead as Opposition Leader.

* A higher Better PM score for Albanese than any acheived by Bill Shorten against either Morrison or Turnbull (Shorten sometimes recorded higher against Tony Abbott).

* The equal second largest turnaround against a PM from poll to poll.  In this regard it trails only a 23 point turnaround against Paul Keating (for the 1993 shocker budget) and is equalled by an 18 point turnaround against John Howard in the above-mentioned Feb 2002 rogue poll.

This is also an unusual case of a PM trailing on Better PM when not far behind on 2PP.  There were a few similar cases with Howard against Beazley, but that was because Beazley was well regarded rather than because Howard was that unpopular.

Newspoll whackings this large tend to generate their own thunderstorms, even after the 2019 failure, and we can expect to see more serious attempts at damage control and a sudden absence of the "just a flesh wound" noises we've seen from the odd anonymous backbencher over this.  Nonetheless there are plenty of prior examples, including John Howard's occasional rough patches, to show that these numbers are recoverable.  It also signals that Morrison is now a political mortal, after an election win that gave him complete authority and a benign six or seven months thereafter.  Whether the rot now sets in and he joins the recent pantheon of dud PMs in polling terms remains to be seen.

2019 In Review

Now time for a belated summary of last year's polling.  As we know the polls got the election held in May seriously wrong, and since then only one poll has returned to the field as far as voting intention polls are concerned.

As a result, despite being an election year, 2019 was the quietest year in terms of total 2PP polls released for decades.  And it continues a rapid downward trend in the number of 2PP polls released in recent years:

This decline had multiple causes.  ReachTEL was an active pollster until June 2018 when the one-year run of its Sky News polling ended.  Commissioned uComms polls using the same platform might have been expected to fill some of the void as the election approached, but publicity of uComms' union links caused media and some activist groups to go cold on hiring them.  As a result, for all the (somewhat justified) robopoll-bashing that occurred in the wake of the subsequent polling failure, there were no pure robopolls in the federal voting intention field in the leadup to the election.

After the election, the polling firms involved in the great 2019 polling failure responded as follows:

Newspoll (now conducted by YouGov, formerly conducted by Galaxy who were purchased by YouGov) did not return until the end of July, and only released eight further polls, bringing its total for the year to 18.  This was the first year since 1991 that fewer than 20 Newspolls had been released.  No federal non-Newspoll YouGov voting intention polls have yet been seen, except for regular Queensland federal polling.

Essential resumed its other polling including leadership polling, but has declined to release voting intention polling for reasons that I find unsatisfactory to the extent that they have even been explained at all.  There has been no way to benchmark poll-to-poll changes in its other figures and say whether they might be caused by a randomly friendly sample for one side or other.

Ipsos were publicly dumped by Nine (owner of the former Fairfax presses) although it had been common in the past for these media companies to take long holidays from polling after elections anyway.

Morgan has not resurfaced.  In the 2016-9 cycle it had polled mostly in the election leadup.

Thus the total number of 2PP polls for the year was a mere 43: ten Newspolls before and eight after the election, three Galaxys before, four Ipsos polls before, ten Essential polls before and eight Morgan polls before.

Of the 35 polls released before the election, the government's best 2PP was 49, which occurred ten times from five different pollsters (most of these in the ridiculous herdfest in the final weeks of the campaign).  The worst was 45 in an Essential one-week sample and in one Morgan that was released post-hoc.  The average of the released 2PPs was 47.8% for the Coalition.  The average using my last-election preferences aggregate (with weighting for past pollster performance) was just 47.3%.  We now know, of course, that all of this was wrong.

Since the election, Newspoll released six polls by its old online/robopoll hybrid (considered discredited by this result - see this article for its full history) and two by its new purely online method.  The government has won seven 2PPs with one tied, and with a low of 50% 2PP and a high of 53.  Its average has been 51.3%, basically the same as the election outcome.  This gives the government an average for those weeks in which polling was active in the year of around 49.6%, but that is meaningless given that we know polls early in the year were wrong.


Scott Morrison averaged a net Newspoll satisfaction rating of 0.5 for the year (up 0.1 points on 2018), making him the first PM to get through a whole calendar year with a positive average since Kevin Rudd in 2009.  (The eight years in between include four with at least one change of Prime Minister).  Morrison averaged -1.5 before the election and 3.5 after the election, the latter thanks mainly to a career-high +15 after winning the election.  Had Newspoll returned to the field faster, Morrison may have polled even higher ratings.  His lowest in 2019 was -9 in the first purely online Newspoll.

Bill Shorten averaged -13.7 prior to losing the election (range -8 to -18) while Anthony Albanese averaged -0.8 since taking over (range -7 to 7).  However, if the poll was either oversampling or overweighting pro-Labor voters before the election, then its ratings for Morrison were probably too low and those for Shorten were probably not low enough.  The reason for this is that Coalition supporters will always rate Coalition leaders well compared to Labor supporters, and Labor supporters vice versa for Labor leaders.

So what happened to all that?

At the end of last year, the government appeared to be in dreadful trouble, trailing 54.2-45.8 in my polling roundup and in a position slightly worse than any government had won from.  A lack of a prior win from such a position didn't mean it couldn't be done, but there didn't seem any reason to suspect that it would.  So what happened to all that?

Nearly half the 5.7 point gap between the government's aggregate polling at the end of last year and its final result was caused by improvement in the government's polling before the election (2.6 points), and about half was caused by error in the final polls (2.9 points).  A very small portion was caused by aggregation issues, mainly in the form of preference-shifting.  It's very possible that the government actually improved by more and that the polls became more wrong during the election leadup and campaign than they were at the end of last year, but this is basically untestable.

Polls found that the government improved firstly over the Christmas break, coming back in early 2019 in better shape than in late 2018, and secondly after the Budget.  Budgets on average slightly harm government polling, but this year's was rather well-rated, and there was an immediate lift of 0.6% in aggregated polling.  At the time it was not clear if this minor gain was real or just noise, but the government's polling then continued to improve slowly in following weeks too.

As of last year's review, I didn't think it was impossible for the government to be re-elected, but I was pretty dismissive of its chances, and at no stage afterwards did I think the government was more than about a 25% chance based on the historic relationship between polling and results.   This was a mistake for at least one reason in the final three weeks specifically, because the degree of herding of the polls meant that the campaign polls were not independent of each other, and hence that an average of them could not be assumed to be as reliable as in the past.  Also, some of Labor's existing policies as of the end of 2018 turned out to be a lot riskier than had been apparent (in part because some of them had been tried in 2016 without doing any harm, seemingly because voters weren't really looking at Labor as an alternative government then.) 

Over time I will be looking at what the 2019 upset has done to the relationship between historic polling and election results.  However, what is left of the ability of polling to explain variation in results may not be very useful for the next election specifically, because there is a probably once-off risk that pollsters will overreact to the 2019 failure, or underreact out of fear of overreacting.  For this reason, when it comes to the 2022 election, unless polls are really lopsided I probably won't be saying that I expect a given side to win.  I'll just be saying things like that if the polls are more or less right, the government could expect a certain seat result, all else being equal.

1 comment:

  1. Great chart on the decline in the number of published polls