Monday, February 19, 2018

Poll Roundup: Barnaby 1, "Momentum" 0

2PP Estimate By 2016 Preferences: 53.3 to ALP (same as end of last year)
With historical One Nation adjustment: 52.8 to ALP
Labor would win election "held now"

With six new federal polls released in 2018 so far it's time for another federal Poll Roundup, a series which I issue roughly every second Newspoll.  As noted in a piece I did about second-term Opposition Leaders, the year began with widespread predictions of a painful year for Opposition Leader Bill Shorten and many people reading a lasting Coalition recovery into, well, a single Newspoll really.  However it has taken the Coalition just two weeks to hit another speedbump in the form of a multi-faceted scandal involving Nationals leader Barnaby Joyce.  The scandal, involving both his relationship with a staffer and various arrangements arising from it, escalated into an exchange of insults between the Prime Minister and his Deputy late last week.

In all this Joyce's behaviour has been that of the ultimate entitled politician, who has even claimed that he himself would never comment on Liberal Party leadership matters when in fact he's been doing so, and frequently, for years. Some of Malcolm Turnbull's comments in response to Joyce's situation have agitated Joyce into retaliating and encouraged Tony Abbott supporters to do their usual thing.  However, the idea that Joyce's lack of moral standards and spectacular political hypocrisy should cause the Liberals to replace their own leader with someone better able to work with such a person is ridiculous.

Polling And Aggregation Changes

There has been a significant change over the break with Essential switching from weekly to fortnightly.  This will mean less frequent polling data and that in some weeks there may not even be any federal polls, but it will also be a lot less work for my aggregate as I will no longer need to be alternately deleting and reinstating successive Essentials to prevent the same datasets from influencing the total twice.  Also following the acquisition of Galaxy by YouGov, the pre-existing YouGov-Fifty Acres poll (which persistently produced extremely strange 2PP results) has yet to reappear, and I would not be too surprised if it has gone.  Some people have been confused about this and have thought the characteristics of the YouGov-Fifty Acres poll would now carry on into Galaxy and Newspoll, but this is not the case - YouGov's ownership of Galaxy has so far had no impact on its polling methods.

There has also been a significant change in polling aggregation land with Bludger Track going over to a hybrid measure of last-election and respondent-allocated preferences.  I won't be doing that here, partly because of an aim to keep the maths of my aggregate simple for easy calculation and checking, and partly because, historically, differences between last-election preferences and respondent preferences often close up as election day approaches.  As a result, readers can expect there to now be a substantial average difference between the two aggregates (currently a point).

What I am doing, because of the evidence that the One Nation preference flow in the few seats they contested in 2016 was unrepresentative, is also publishing an adjustment figure for those who want to measure One Nation preferences at historic levels.  I'll also be using the adjusted figure myself for seat projections.  Another part of the difference between respondent and last-election preferences is that it may be that 2% or so of voters nationwide are favouring Cory Bernardi's Australian Conservatives.  Said party preferenced the Liberals extremely strongly at the Bennelong by-election.  As Australian Conservatives are a new party, should any useable measure of their national support become available in polling, I may in future carve off part of Others to them and use their flow from the Bennelong by-election as the next best thing going to a "last election" flow.

Voting intentions

The first Newspoll this year came out at 52-48 to Labor, ending a run of 14 consecutive results of 53% or more.  This had been the third longest such run in Newspoll history, after runs of 29 such results under Julia Gillard in 2011-2 and 25 such in the last year of the John Howard era.  One Nation dropped from 7% to 5% with the party seemingly in decline after not living up to the hype in two state elections last year.  However, this week's Newspoll came out at 53-47 with the points lost to One Nation returning to sender (indeed One Nation also gained a further point from Others).  That was the Coalition's 27th consecutive 2PP loss under Turnbull, compared to the famous 30 in a row under Abbott, and the probability of Turnbull jagging a 50-50 in the next three polls would now have to be very low indeed.  The other polls out have been three Essentials (a 53-47 followed by two 54-46s), and also one ReachTEL which was 52-48 by respondent preferences, but I got 54.2 to Labor by 2016 preferences.

So really the evidence for the Coalition's New Year polling momentum that spawned numerous op eds consisted of voting intention movements well within movement of error in a solitary Newspoll and not backed up by any other pollster, plus leadership polling fluctuations that are generally much more volatile.

Moreover, the last four Newspoll 2PPs in a row have all been very generous to the Coalition by last-election preferences, with the readings of 47, 47, 48 and 47 coming from primaries that would normally have given 46.3, 46.3, 47.3 and 46.3 respectively.  While these differences are individually within the range of rounding and state-based preference calculation issues, I do wonder if the calculation may have been changed to adjust for One Nation issues.  For now I've aggregated these as three 46.6s and a 47.6 pending clearer evidence on this.  Another curiosity of this Newspoll, as noticed by @sorceror43 on Twitter, is that it was the 11th in a row where the combined ALP/Green primary vote was 47%.

Overall my aggregate started this year with a mini-blowout when it only had the first Essentials and the ReachTEL in the mix, but it has settled down to a current last-election value of 53.3 to Labor (52.8 with One Nation adjustment), which is where it ended last year.  Here's the smoothed tracking graph:

Since its initial blowout phase, which ended in early 2017, the aggregate basically hasn't done anything sustained in either direction.  If it isn't a random walk, it may as well be.  Absent of something causing either a sustained or a quick and major shift in one direction, the blue line will simply random-walk to a defeat, probably on a similar moderate scale to the 1983 and 2007 losses.

It may well surprise people that this week's Newspoll wasn't worse after all that has happened with Joyce.  However Joyce is a National, and Nationals as such only make up about 3-4 points of the Coalition's primary vote polling.  One would expect Joyce's behaviour to rub off on his own party, but it won't necessarily rub off on the Liberals and the Queensland LNP.


The leadership figure of most interest this fortnight involves Barnaby Joyce, with 65% wanting him to resign from the leadership, with 23% wanting him to stay and the other 12% uncommitted.  Among Coalition supporters, 58% want him to go and 35% want him to stay (it was initially misreported that only 33% wanted him to go).  In total 21% want Joyce to quit parliament immediately,  15% want him to quit as leader and retire at the end of his term, and 29% want him to quit as leader but continue as a backbencher.  There are many more damning findings.

A ReachTEL of the seat of New England had Barnaby Joyce's primary vote crashing to 43% in a hypothetical contest including former MP Tony Windsor (26.1%), which came out to 52-48 to Joyce by respondent preferences.  However Joyce's vote would be more like 45% once undecided voters were redistributed, and that's only a 7% swing against his 2016 election result, most of which has probably gone to One Nation (who didn't contest the seat).  Bearing in mind also that polls had New England as a close contest in 2016 when it eventually wasn't, I think not that much can be read into this result yet.  The poll also featured a stunning demonstration of the extreme uselessness of polls that ask an elector whether something makes them "more likely" to vote a certain way, with 25.4% lying that Barnaby Joyce's affair had made them more likely to vote for him.  It is really time outlets stopped bothering with a form of polling that just encourages partisan responses.

In defence of Joyce various myths are circulating about him having saved the Coalition's bacon or at least its majority at the last election.  In fact, while the Nationals gained a seat at the 2016 election, that was at the expense of the Liberals.  While the Nationals did avoid losing any seats, only two of their MPs were on margins below 6%, compared to 25 other Coalition MPs.  One of the Nationals on below 6% was an LNP MP caucusing as a National (public awareness of which LNP members are Liberals or Nationals is not that strong).  Most of the Nationals-as-such occupy safe rural seats where the only threats are sporadic independent uprisings.  This is less the case for LNP Nationals, but some of those including George Christensen and Ken O'Dowd copped quite nasty swings against them and their seats are now very marginal.

In Newspoll, PM Malcolm Turnbull is at a net rating of -20 (34-54) this week (down from an unusually high -13 last fortnight) while Bill Shorten again had exactly the same figures as Turnbull.  These were initially (and at the time of writing still are) graphed by the Australian at -18.  As noted in the opposition leaders piece, Shorten has now been in negative double digits continuously since March 2015, a record for any Opposition Leader or PM.  Meanwhile, Turnbull's lead as "better Prime Minister" is back down to seven points (40-33) having climbed to 14 points in the previous poll.  While that should have pulled the rug out from under the "momentum" and "Shorten pain" narratives, it is still worthwhile pointing out that Shorten has never led Turnbull as better PM, although Shorten led Abbott slightly more than half of the time.

I thought I would have a look at some multiple regressions for Turnbull's lead as better PM over Shorten, compared to Abbott's.  In both cases, virtually all variation (93-95%) is explained by a combination of the 2PP polling, the PM's netsat and the Opposition leader's netsat (in that order).

For Abbott the equation was:

Abbott better PM lead = 58.4+0.49*Abbott netsat - 0.96*Labor 2PP - 0.18* Shorten netsat

For Turnbull:

Turnbull better PM lead = 73.5+0.45*Turnbull netsat - 1.13*Labor 2PP - 0.36*Shorten netsat

On this basis, even taking into account differences in ratings for the individual leaders and shifts in two-party preferred polling, Turnbull's better PM ratings vs Shorten have still been much better than Abbott's.  Had Abbott still been PM, Shorten would probably have led him as "better PM" on every Newspoll since the last election.  (Some of this could be seen as a smoke signal that the post-election 2PPs are wrong, but in fact this Turnbull effect was stronger before the 2016 election than after it.)

Essential last week gave Malcolm Turnbull a net rating of -3 and Bill Shorten a net rating of -13, with Turnbull leading Shorten as better PM by 15 points.

Bonkban and other polling

The other Newspoll question of note concerns Malcolm Turnbull's "bonkban" against Ministers having "sexual relations" (whatever that is) with their own staff.  The Newspoll question goes further and asks:

"In the United States politicians are banned from having sexual relations with any employee that works under their supervision.  Would you be in favour or opposed to a similar law being introduced in Australia?"

This question then finds a 64-25 support rate for a measure that is much stronger than the "bonkban" in two regards - firstly it applies to all politicians, not just Ministers, and secondly it would carry the force of law, not just the force of dismissal from Cabinet.  However, I don't like the question design as the information that such a ban exists elsewhere is an argument in favour of supporting it here.  It would be better to ask:

Would you be in favour of or opposed to a law banning politicians from having sexual relations with any employee that works under their supervision?

On some other matters, Essential finds voters thinking the Adani coal mine will both create jobs and take away jobs.  It also finds substantial improvement for Labor on some party trust indicators, including "political leadership" (which is interesting in light of Shorten's poor scores on this front.)  Comments are welcome on any other issue polling out there that anyone has found of interest; I need to get back to finishing an overdue piece on the mess that is the SA state election!


  1. "It may well surprise people that this week's Newspoll wasn't worse after all that has happened with Joyce. However Joyce is a National, and Nationals as such only make up about 3-4 points of the Coalition's primary vote polling. One would expect Joyce's behaviour to rub off on his own party, but it won't necessarily rub off on the Liberals and the Queensland LNP."

    I would have thought that the average voter would struggle to explain the difference between the Liberal Party and the Coalition, and just see it as a bad Deputy PM reflecting badly on the Government.

    Then again, I'm from Queensland and in my 20s, I suppose the Lib/Nat distinction would be much better understood in other states and other age groups.

  2. Unless there is a 'circut' breaker' these poll results will continue to the election. Neither Turnbull or Shorten are capable of delivering that.


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.