Sunday, February 8, 2015

Liberal Spill And Poll Roundup: The Trouble With Political Jokes

2PP Aggregate: 55.8% to Labor (+3.2 points in three weeks, highest reading of term)
(updated following Newspoll and Morgan)
Labor would win election held now with a massive majority.

The Coalition is now in its worst polling position of the current term
(NB Scroll to bottom of article for updates for polls and other events added from Sunday 8 Feb on)

I've just returned from a week and a half on remote field work, and was going to put out my first poll roundup for the year on Tuesday.  However, all sorts of things might happen by then, so the data might very well be redundant.  I've decided to catch up on the mass of recent polling data first, and then follow up on the early week polling when it comes out.  But first, time for a look back at how (in my view at least) we got here.  My own contribution to the narrative-writing about what is going on now is rather long, so some may wish to skip the next two sections.

Accidental PM

Once upon a time one Tony Abbott was a mostly able frontbencher, respected former party staffer and prominent monarchist whose career was however rather incident-prone.  These incidents stemmed partly from Abbott's frequent expressions of values from a 1950s time warp on issues such as the monarchy, abortion, homosexuality and the place of women, and also partly from a colourful past and a proneness to sporadic lapses of self-control.  They did not prevent him from being a major party figure and source of Coalition political thought.

However, because of Abbott's hardline and erratic reputation, the idea of him becoming leader was very difficult to take seriously.  He was widely viewed as "unelectable", and his elevation to leader was a cause of widespread mirth among ALP supporters.  He became leader by one vote when Malcolm Turnbull, spectacularly damaged by an unsound attack on then well-regarded PM Kevin Rudd in the Ozcar affair, became unpalatable to the party base by telling the party to support Labor's Carbon Pollution Reduction Scheme.  (Rebellion came both from those who could not stomach this decision ideologically and those who just thought it was bad politics.) Even then, the more moderate Joe Hockey was supposed to be the man for the job - except that when put to the test Hockey could neither argue a clear position on emissions trading nor navigate the murky tactical waters of a three-way spill.

The best Liberals could reasonably hope for when Abbott ascended to the leadership was that he would be a caretaker who would keep the base happy and avoid losing the next election by too much, by which time some of the gloss would have gone off Rudd's popularity and the damage done could be healed.  However not long after Abbott's ascension Rudd granted the Liberals two big gifts - firstly the Mineral Resources Rent Tax which caused major mining interests to start campaigning against Rudd, and secondly the brand-damaging dumping of the proposed Carbon Pollution Reductions Scheme in the face of Senate opposition.

Labor then gave Abbott an even bigger gift by using a sharp decline in Rudd's polling as a pretext to roll him and install Julia Gillard.  Labor managed to just win the election but spent the whole of the 2010-3 parliament hamstrung by public perceptions of the hung parliament thus created, and by infighting.  There were those who considered Abbott some kind of genius as Opposition Leader for bringing Labor to its knees in this manner, but that's giving both sides far too much credit.

Abbott ticked some very modest boxes required to keep the Coalition competitive - the idea that an Opposition should actually oppose, for instance. Then, under a level of political pressure that most past governments would have considered trivial, Labor suddenly self-destructed as a result of a combination of personality flaws and internal greed and power games.

Abbott became Prime Minister in late 2013 but nothing in subsequent polling has suggested the public chose him based on any real positive merit or enthusiasm.  His own ratings remained in positive territory for just two months and his party kept the 2PP polling lead for only three - the worst start by far for any new government elected from Opposition.  The theory that the election of the Coalition was mainly about policy or economic management issues making Labor unelectable has since been found wanting at every available polling-based test.

The theory that Abbott mainly won the last election because the voters saw the Rudd/Gillard/Rudd soap opera as even worse (or as eviction-worthy whatever the alternative), on the other hand, continues to pass with flying colours.  Abbott only made it to the Lodge because the 2010 result was well above expectations, and then the party led in polling almost throughout the 2010-3 term, making him immune to deposal.  In the role, Abbott has generally shown more restraint and wasn't quite the instant joke that detractors expected, but he has led the party in a very unpopular policy direction.

Finally, the recent decision to confer a knighthood on Prince Philip was the "Mad Monk" of the old days and then some. Descriptions of this incident as a mental aberration are too kind.

There is an old line that the trouble with political jokes is that they get elected.  That saying doesn't say what happens next.

Abbott Camp Arguments Are Weak

Listening to the early language from Tony Abbott and his supporters about the upcoming spill, there are two common themes.  The first theme is that "we are not the Labor Party" (complete with some rather nice lines about politics becoming Game of Thrones) and the second is that the party room would be removing the Prime Minister elected by the people, rather than letting the people make that decision themselves.

The ALP's leadership chaos during its previous terms arose partly because the case for removing Kevin Rudd was never adequately made and partly because Rudd decided to stick around.  Having been removed after a brief period of unpopularity and with his party competitive in the polls, Rudd remained a viable contender.  In Abbott's case he is so unpopular, and his party's voting intention figures so bad, that his removal would spell the end of him as a viable contender for his party's leadership.  He might well resign from parliament within months, but if he did remain he would not have a legitimate grievance to justify trying to get the job back, and would have little sympathy if he tried.

That's not to say a transition to a new PM will be smooth.  It would be likely - especially if Malcolm Turnbull emerges as the new leader - to come with a substantial Cabinet reshuffle and an exodus or else purging of ministers appointed by reasons of loyalty to the Abbott view of the world rather than talent, or else (like Hockey) too bound up in Abbott's failures.   There is the possibility the new leader might not be up to it either.  But the hallmark of the Labor administration was not simply that it had a brief period of leadership tension, but rather that leadership issues existed on and off for three years and developed into soap-opera proportions.

The popular-election argument also doesn't work.  Kevin Rudd won election from Opposition as a popular leader defeating a respected incumbent on a wave of positive sentiment.  A brief falling-out with the public after a career of excellent ratings did not eliminate a sense of public ownership of his career and that the voters had been robbed by not being allowed to make their own ultimate decision.  Abbott, however, has never had the same public loyalty and polling suggests the voters have made up their minds about him already, are unlikely to like him in the future, and will not be at all aggrieved if the decision is made for them now.

Spill history implies doom for Abbott

Prior to this one, I have been able to find 13 cases of Prime Ministers or Opposition Leaders facing spill motions or leadership contests of some kind.  The affected PMs were Gorton, Fraser, Hawke, Rudd and Gillard.  Affected opposition leaders were Snedden, Hayden, Howard, Hewson, Crean, Beazley, Nelson, Turnbull.  Of these, four (Rudd, Howard, Beazley, Nelson) were removed immediately.  Eight (Gorton, Hawke, Gillard, Snedden, Hayden, Hewson, Crean, Turnbull) survived the initial spill but were removed in a subsequent spill or had resigned by the time of the next election.  Only one (Fraser) was still leader at the next election, which he lost.  So, assuming the spill motion yet goes ahead and attracts some kind of support, there is no precedent for Abbott winning the next election from here, and only one precedent out of thirteen for him even still being PM.

Abbott has defied history before - winning office from far worse personal ratings than any previous Opposition Leader - but this one coupled with a very poor 2PP polling record seems much more predictive.  That said, at this stage we're still talking about just an expected spill motion.  There might be some other resolution by Tuesday.

Those who think that rolling the leader necessarily spells political death for a party (an argument being advanced for avoiding it) should be aware that this is not well supported by the historical record.  Of twelve leaderships that ended during the term in which they had been the subject of a spill motion or contested ballot, in six cases (Snedden to Fraser, Hayden to Hawke, Hawke to Keating, Hewson to Howard via Downer, Beazley to Rudd, Rudd to Gillard) the party in which the spill happened still won the next election.  Excepting the last, it's hard to make much of an argument that any of these changes were a bad idea.

Polling: voting intention

Since the resumption of polling this year there have so far been twelve national 2PP polls.  These were:

* Two startling Galaxys in the last two weeks, each with a 57:43 headline to Labor, These are the worst polls for the Coalition in its term, excluding a one-week portion of a Morgan sample that Morgan published separately because it was so different to the other week's.   It is interesting that Galaxy's readings have turned so vicious for the Coalition in the last three federal polls.

* Two ReachTELs in the last two weeks with a 54:46 lead to Labor (which I aggregated as 53.8:46.2 based on decimal primaries) and 55:45 (left as is).

* One Ipsos poll with a 54:46 lead to Labor, off primaries that imply the lead was more likely in the low 54s than the high 53s (I aggregated it at 54.2).

* Two Morgan multi-modes with Labor 2PPs of 53 then 55.5 to Labor by last-election preferences  (54.5 and 56.5 by respondent preferences).  This series skews to Labor by about 1.5 points, so the 53 was actually a rather benign start-of-year result.

* One Morgan phone poll with a small sample size and a 53.5 to Labor result.  Note that Morgan phone polls do not skew to Labor as Morgan multi-mode (SMS/F2F) polls do.

* Four Essential polls with Labor 2PPs of 54-53-54-54.

The first Newspoll is coming this weekend (I would think it might be released on Sunday night instead of Monday night in the circumstances) and we should also get Morgan on Monday and Essential on Tuesday this week.

Aggregating the polls

One issue in aggregating federal polls at the moment is the possibility that Ipsos has a house effect. The first three national Ipsos phone polls have all been more friendly to the Coalition than my aggregate at the time, by exactly one point on average in my 2PP estimates as included in my aggregate (slightly more than a point if you just use the published 2PPs).  There are two possibilities here - firstly that Ipsos actually has a significant house effect, and secondly that it doesn't and that these three results are largely a matter of chance.

Establishing which it is is actually a tricky exercise because we don't know the a priori chance that a new pollster has a substantial house effect in favour of the Coalition.  The chance that Ipsos would produce this kind of leaning if it had no house effect is only about 10%.  However if we assume that a typical new pollster is more likely to have no house effect than one in Labor's favour, then our estimate of the chance that Ipsos has no significant house effect will increase.

I'd like to see a little more evidence on this score, so rather than declare that Ipsos has a house effect at this early stage, I've addressed the apparent lean of the whole pool of pollsters (after adjusting Morgan) through the global house effect correction built into my model.  Until the evidence on Ipsos is clearer, Labor gets 0.1 points added to its total, and I was very close to making that 0.2.

This brings Labor to 55.6% 2PP (55.0% at the end of last week with Galaxy doing the rest), which is Labor's best position on my aggregate in this term.  The smoothed tracking graph after ReachTEL can be seen at the top of the article.  The blowout in the last two weeks may look worse than it was because the previous two weeks saw mild readings based on very limited data from Morgan and Essential only.

Changes in respondent preferences

But maybe the Coalition are doing even worse than 45-55.  We've just seen the Queensland election in which previous-election preferences proved, for once, to be extremely unreliable.  The Queensland election was a special case in this regard because of the dramatic swings in primary vote and the nature of the government thrown out at the previous election, but even that did not explain everything.  After the last federal and Victorian state elections (in both of which the reality of preferencing was roughly in between previous-election and respondent-allocated) it's worth paying some cautious attention to what respondent-allocated preferences are doing.

The following graph shows a rolling five-poll average of the difference between Morgan's respondent-allocated 2PPs and their last-election 2PPs since the last election. A positive difference is one in Labor's favour.

At the start of the government's term there was no real difference between respondent-allocated preferences and preferences from the election just held.  A difference appeared gradually, reaching about 0.8 points in April 2014 just prior to the infamously unpopular Budget (poll 17).  At this time the difference quickly jumped to around 1.5 points.  It had retreated to about 1 point (poll 31) in November 2014 but might be picking up slightly again.  If the pattern of the 2013 election repeats and there is a further shift in preferencing behaviour (just not as much as respondent preferences imply) then the current 55.0 to Labor could be worth, say, 55.6.

Leadership Polling

1. Leader approval:

So far this year we have had the following leadership results:

* In mid-January, Morgan (phone) had Abbott on a -15 netsat (approve 37, disapprove 52).  This result was benign compared to those to follow.  Morgan had Shorten on -3 (37-40).

* In very late January, Ipsos had Abbott on an awful -38 net satisfaction rating (approve 29 disapprove 67).  We have little standard of comparison but Ipsos' methods are largely calibrated to those of the old Nielsen series, in which readings of this kind were exceptionally rare.  Ipsos had a +10 net approval for Bill Shorten (48-38), among the best of his career.

* ReachTEL had Abbott's performance rated as Good or Very Good by 21.6%, Satisfactory by 16.7% and Poor or Very Poor by 61.6% on 27 Jan.  On 5 Feb these figures were 21.1, 16.2 and 62.7.  These were Abbott's worst ReachTEL results as PM.  ReachTEL's figures tend to look unflattering to Bill Shorten because of the use of the middle Satisfactory option - in late January 27.1% rated his performance as Good or Very Good compared to 38.3% Poor or Very Poor, and in early February 25.1% and 39.3%.  The late January result was Shorten's best for a year.

2. Preferred PM (Abbott vs Shorten):

Morgan phone had a 43-41 lead to Shorten in mid-Jan.  Ipsos had a whopping 50-34 lead in late January and Galaxy had an even larger 44-27.

Essential had head-to-head attribute comparisons.  Since December Abbott's attribute scores have worsened on everything but "aggressive" (down 4) and "more honest than most politicians" (unchanged).  Most of the scores are also worse than they were after the Budget.  Shorten's scores had also mostly worsened, but only by 1.3 points on average, compared to 3.4 points for Abbott.

The comparison table showed Shorten outrated Abbott on every question except "More honest than most politicians", and that one would be just because fewer voters had opinions one way or the other re Shorten.  On average Shorten outscored Abbott by 14.7 points.

3. Preferred Liberal Leader

* Morgan phone in mid-January had a smorgasbord poll with Turnbull (36) preferred to Bishop (26) and Abbott (14) with 13 for others and 11 undecided.  Abbott (30) lead Bishop (28) and Turnbull (26) among Coalition voters.  (Bill Shorten continued to weakly lead Tanya Plibersek as preferred ALP leader 25-18.)

* ReachTEL on Jan 27 had Turnbull (44.6%) preferred over Bishop (30.5), Abbott (18.1) and Hockey (6.7).  Among Coalition supporters, ReachTEL had Abbott (39.6) preferred over Bishop (27.9) and Turnbull (27.1).  Labor and Greens supporters strongly preferred Turnbull.

* ReachTEL on 5 Feb had Turnbull leading Bishop 56.5:43.5 if those are the only two options, but Bishop led 54.4:45.6 among Coalition supporters.  There was a twelve-point gender divide on this question.

* Essential in early February found Turnbull (24) preferred as leader to Bishop (21) and Abbott (a mere 11) with 5 for Hockey, 2 for Morrison, less than 1 (ouch!) for Pyne, and 13% "someone else" and 24% "don't know".  Essential differed from the others in showing Abbott trailing even among Coalition voters - 26% for Bishop, 24% for Turnbull, 23 for Abbott.

Newspoll are reportedly currently polling Abbott vs Turnbull, Abbott vs Bishop and Turnbull vs Bishop.

(A note here: I've received some flak for drawing attention to Essential's whopping don't know rates, suggesting that they represent a method issue specific to online polling and not a flaw.  I maintain that it is both a method issue and a flaw.  In an online survey a don't-know option must be provided on the same level as other choices, but this makes it very easy for the respondent to quickly whack don't know and move to the next question, especially if their motive in filling out the survey is a shopping credit reward.  Given that voters will often pick an option when "don't know" is not explicitly given as an option, I believe that the online format is conducive to voters with genuine leanings sometimes failing to give their view.  Those leanings are of interest because voting decisions do not solely rest on those opinions that a person is most confident about.)

4. Hypothetical 2PP with different leaders:

In the comparative leaderships department, ReachTEL had 54:46 to the Coalition if Turnbull was leader and 51:49 for Bishop, in a poll with an overall 45:55 result under Abbott.  Galaxy had 49:51 with Turnbull, 47:53 with Bishop and 43:57 with Abbott.  So on average these two anticipate a bounce of 7.5 points for installing Turnbull and 5 points for installing Bishop.  However, the track record of these sorts of hypothetical polls is that they are poor at predicting anything other than the immediate bounce.  We saw this in 2013 with a slew of polls suggesting Rudd would do very well as Labor PM - he more or less matched the expected results initially but was way short of them at the election.

5. Other

Ipsos polled belief that leaders would lead their parties to the next election.  Voters did not believe Tony Abbott would (31-60) and did believe Bill Shorten would (70-16).

Polling on the Prince Philip knighthood has been unsurprising with heavy rejection being found by Ipsos (15-74), ReachTEL (12.1-71.5), Essential (14-69) and Galaxy (14-70).  Both Essential (26-46) and Ipsos (25-55) also found voters rejecting the reintroduction of knight and dame honours, though in the former case not by much more than before.

The most recent Galaxy found 55% believed Tony Abbott should resign as Prime Minister, compared to 35% who thought otherwise.


I'll update this article following polls over the next couple of days, including Newspoll.  The headline voting intention figures will create much interest there, but if they are benign that does not necessarily mean Abbott is off the hook.  Around the time of Labor's 2012 spill, polling for Labor briefly picked up in the expectation that the party was on its way to reinstalling Kevin Rudd, then crashed again soon after.  Unless the Newspoll is at least as bad as the Galaxy, then I think the real interest will be Abbott's personal ratings.

A widespread view is that this spill is round one and that potential challengers are testing the water to see how much of a mood for change exists and cause damage ahead of an eventual round two.  However with the spill motion on the table, it's possible events (including more bad polling) between now and the spill being voted on, could see Abbott's position collapse right away.  If it is a two-stage strategy then the challengers had better get a move on to ensure the issue is resolved in advance of the NSW election.  While it is generally assumed the Liberals have re-election in NSW in the bag, that is not a safe conclusion if the preferencing shift seen in Queensland is repeated and if federal leadership issues are still on the table.

Newspoll Update

The current Newspoll is appalling for the government, but it probably could have been even worse.  Firstly the government trails 43:57 on the headline 2PP, though the primaries are just a smidgen friendlier than Galaxy's.  Both the Keating government in Keating's very early days and the Howard government once in 2001 were able to come back from exactly that figure to win without changing leaders.  No-one has ever won without leadership change from further behind in Newspoll, but the Menzies government won from further behind on old Morgan Gallup polling in its 1951-4 term.

Tony Abbott recorded a net satisfaction rating of -44 (24-68).  These figures are identical to the worst ever polled by a PM who was returned at the end of that term (Keating).  Worse figures were recorded only by Keating four times in August-October 1993 (including two -57s) as well as a single -45 (23-68) for Julia Gillard in September 2011. Only three Keating Newspolls in the term in which he lost had higher dissatisfaction ratings, making Abbott's the equal fourth worst for a PM ever polled.  They are also Abbott's personal worst (beating even the -36 he recorded once as Opposition Leader).

Bill Shorten recorded a net satisfaction rating of +2 (42-40).  Ho hum (EDIT: It is his best since mid-May, but the level of overall variation has been very low).  Shorten also recorded a 48-30 lead over Tony Abbott as better Prime Minister.  This was the second-highest lead ever recorded by an Opposition Leader, falling two short of the record held by Alexander Downer (which tells you much of what you need to know about what a shaky indicator this is).  Only Keating as incumbent PM recorded worse raw better PM scores (five times, but one was when he'd just arrived in the job and the other four were in the term in which he was defeated.) Only Kevin Rudd has polled a better PM score higher than Shorten's 48 as Opposition Leader.

Newspoll also polled three Liberal head-to-heads and found Turnbull leading Abbott 64-25, Bishop leading Abbott 59-27 and Turnbull leading Bishop 49-38.  Given the common preference of left-wing voters who would never vote Liberal in a fit for Turnbull, these figures actually imply that Bishop is a very viable leader in the mind of those whose views should actually matter.

In an article in mid-2014 (Turnbull PM: Not Likely Any Time Soon) I ran through a list of reasons why I thought Turnbull wasn't about to be installed, along with all the reasons why he was being considered at that time.  Of the reasons against, here's an update on some of them.

1. The government's polling isn't actually that awful yet: No longer applies; the government's polling is now clearly awful.

2. The government's polling must be seen in context: No context consideration applies.  The government's polling results not from tough decisions but from stupid and out-of-touch ones.

3. Turnbull is not the preferred leader of Coalition supporters: This probably remains true, but Abbott may not be anymore either. (That said, Newspoll shows him preferred to either Turnbull or Bishop head to head in this regard.)

4. Turnbull's polling reflects rose-coloured glasses.  This remains true, but does not apply to Bishop.

7. The timing is just wrong.  It may still be too early for a switch, but is fast becoming less so.

The flow of recent attempts to predict tomorrow's spill motion has been that Abbott is expected to survive with maybe only a couple of dozen votes against him.  Even a Newspoll as bad as this may not be enough to swing the outcome so close to the meeting, but it could increase the damage.  The best Abbott can hope for is that the number of votes against him is embarrassingly small and the spill attempt gets seen as just the work of a small number of panicking backbenchers.  I doubt he'll get away that easily from this, and the kind of language he is using in promising to do better suggests he is seriously worried at least that he may not win by enough.

Abbott Survives Spill, Margin Not Enough

The spill motion has been defeated 61-39 but the margin would be disappointingly high for the PM - this is not just a moderate backbench panic but a much more serious revolt.  Historically this level of dissent is not the end of the matter, but rather the beginning of the opposite end.  We also saw today probably the most abject grovelling from an incumbent PM in the history of Federation.

After the result a new Morgan came in with a pretty middle-of-the-road (given Morgan's house effect) 57:43 to Labor (57.5 by last-election preferences).  The poll was taken over the last two weeks so some of the data would have been pre-spill.

Tuesday: Essential has displayed its usual indifference to large moves picked up by other polls, again returning 54:46 as a rolled average for the last two weeks.  Tony Abbott has an Essential netsat of -33 (29-62) which is his worst since becoming Liberal leader.  Bill Shorten has curiously gone backwards from +6 to -5.  Shorten has an eight-point lead as Better Prime Minister (39-31), which is his highest lead from Essential so far.  Although the raw scores in Essential are less dramatic than from the phone pollsters, the overall patterns are similar.  If we're all done for the week now (with a whopping 11 polls now contributing data) then the aggregate will finish the week at 55.8 to Labor, up 0.8 points on last week's close of 55.0 (adjusted for global house effect correction).  Labor's previous peak end-of-week reading was 54.9.


  1. Any explanation why Essential has this "usual indifference"?

    1. That is a great mystery to many of us. I first looked at it in detail last year ( but I don't think I exhausted all the possible explanations there. Essential uses a two-week rolling average which should make it less dynamic than some other polls, but is not enough to explain how stable it is by itself.

      I think the main contenders are: something in the way the company scales polls, some underlying bias of the company's panel of potential respondents, or something in the way the company chooses respondents from its panel.

      Because Essential chooses respondents from a panel of people who have in common that they like filling out surveys on the internet to get shopping vouchers, there are probably some small biases in its panel group that no amount of scaling can get rid of. If such people, for some reason, had more stable political opinions than others, then that would explain Essential's slowness to follow trends. However it wouldn't explain their poll-to-poll stability even in the absence of any trend.