Thursday, March 29, 2018

Poll Roundup: Number 29

2PP Aggregate: 53.3 to ALP (-0.3 since last week)
Labor would easily win election "held now"

This week's federal polling coverage was dominated by Malcolm Turnbull slipping another Newspoll closer to the dreaded number 30.  Although there may be minor signs of improvement, it would be highly surprising to see the Coalition's 2PP jump to 50% in two or three weeks' time.

Since the last poll roundup, we've had two Newspolls (both 53-47 for Labor), three Essentials (53-54-52 for Labor), and two ReachTELs, one in late February and one today.  The ReachTEL was 54-46 by respondent preferences, but I estimated 55.5-44.5 by last-election preferences, which made it the single worst poll of this government's term in my aggregate (though not by much).  The March ReachTEL was also 54-46 but in this case I got 54.2-45.8 as a last-election estimate.  Oh and there's also, to my great surprise, been a Morgan, and that gets a section of its own below.

The late February polling surrounded the Barnaby Joyce scandal, and now we have a completely different media cycle that's roughly 80% cricket cheating and 15% franking credits, so it's not too surprising that the recent polls haven't been quite so bad for the incumbents.  But the improvement is barely noticeable and not even statistically significant - the government continues to flounder around in the mid-46s on aggregate as it has since last August:

The very likely loss of Newspoll number 30 will create great interest on social media, but no-one much outside some Labor partisans and Abbott fanatics seems to think it is going to blow up into a leadership-losing event (at least not by itself).  It might be interesting to see the reaction if it is also the first of the 30 to have Bill Shorten ahead as Better Prime Minister.  In a Saturday Paper piece on the moral case for booting Turnbull upon him reaching his 30th loss, Chris Wallace makes a case about how Turnbull's use of the 30-Newspoll argument to get rid of Abbott has a lot more parallels to Turnbull's own situation now than just the number of polls - it was not just a single throwaway line, it was a core argument of unelectability and lost opportunity to improve as the clock to an election ticked down.  But there's one comment in particular I thought I'd pick up on:

"The last time the Turnbull government led Labor in Newspoll on a two-party-preferred basis – July 2, 2016 – was the day of the last federal election. It is the most complete case of buyer remorse in Australian political history."

Perhaps it is, but we don't know that for sure, and one reason we don't is that Newspoll took nearly two months to return to the field post-election.  There have been a number of governments that started losing very soon after the election, but that did get a few leads in the early months.  Labor after 1990 was a good example of this, but only became competitive by changing leaders (see my recent special feature The Keating Aggregation).  The more important issue is that the stability of the new Newspoll means that a party that is trailing will no longer get the one rogue sample in its favour every now and then as used to happen.  Though the polling in those days was much less frequent, my nomination for the most complete case of buyer's remorse would be Menzies after the 1951 election - and he won.

This week's Newspoll primary votes were notable for Labor hitting a term high of 39%.  This also equals Labor's term high from any pollster, as Essential returned two 39% primaries for Labor in August.  The Australian seemed surprised by this, given that its Newspoll showed 33-50 opposition to Labor's franking credits plan, but it doesn't follow just from that issues poll result that the issue is a big vote-loser for the Opposition.  Indeed a lot of the analysis about the impact of the change on Labor's popularity has been superficial.  It has ignored that retirees are more likely to vote Liberal, and also that when there are prominent complaints about retirees getting gouged there will also be some younger voters who decide on that basis that the proposed change is good.  Generational conflict is quite a big thing in Australian politics, with media frequently publishing dumb articles whinging about "millenials" and posters on social media frequently denigrating "boomers".  Successful opposition to Labor's plan will require communicating its impacts on a range of investors and not just those over 65.

Not much to see on the current leaderships front with Turnbull's netsat at -24 last time and -25 this time and Shorten's at -23 then -20.  However, PM Turnbull recorded an equal term-low lead of two points as Better PM last Newspoll, adding only one to that this fortnight.  ReachTEL polled on the question of Tony Abbott returning to the Liberal leadership after the next election, a concept that has been mooted if the Coalition loses, finding 25% support and 64% opposition.  However the reporting of the poll question does not make it clear whether the question was confined to if the Coalition loses.  (Not confining it would skew it against Abbott, since some voters would think he is unsuitable to be PM but would be fine with him being Opposition Leader again.)

There are a number of interesting issues polls doing the rounds at the moment (and also some pretty boring ones, such as those about cricket or, worse still, renewable energy).  In view of limited time and energy I am not going to discuss them right now, but may append some discussion over Easter if time permits.

Bring Out Your Dead!

A very odd development in polling this week was the unexpected appearance of a Morgan Face-To-Face, including results from a previous fortnight of the same.  Fully fledged Morgan face-to-face polls haven't been seen since February 2013.  The series was notorious for being skewed to Labor by 2-3 points much of the time, a skew likely to have been caused by social desirability bias.  We see some potential signs of this in the current poll too - One Nation polls half what it is getting in other polls, while the Greens vote is inflated.

Morgan continued using face-to-face in a hybrid form in its multi-mode polls for a few years after shelving the purebred version.  These hybrid forms showed a watered-down version of the same effect, with their house effect swinging to the Liberals when Turnbull first took over and was briefly very popular.

It's not clear from a sample size of two whether the current incarnation has any 2PP house effect at all.  If it's to be believed, the Coalition has pulled back three 2PP points in a fortnight, which might if true be attributed to Labor's franking credits plan (Morgan suggests the SA election or Adani, but really the explanation is irrelevant because the three-point jump just didn't happen anyway).  The key point here is that the beneficiary is mostly not the Liberal Party but the National Party, up 3 points from 2.5 to 5.5.  Morgan's suggestion that this is down to the impacts of Labor's Adani position in Queensland makes particularly little sense since in Queensland there is no National Party, only the LNP.

One might think that this is no big deal; after all the margin of error for a sample of 1417 electors is 2.6% so so what if a party bounces a few points from poll to poll.  But actually the margin of error referred to by pollsters is the maximum MoE, applying to events with a score of around 50% (such as the 2PP).  For small scores, the MoE is smaller, and the MoE on a reading of 2.5% is only +/- 0.8.  In a remotely random sample, National Party support would only be remotely likely to more than double if it had at least nearly doubled in the wild, which is of course silly.  So as evidence that the 2PP can change (which of course it can) this poll isn't very convincing - it's much more likely the difference between the two face-to-face polls is caused by sampling or scaling issues that probably go beyond mere random sample noise.

There are still more issues with this poll release by Morgan.  Firstly "Poll watchers will have noticed Morgan Polls have not been publicised regularly. Morgan Polls are still conducted weekly, and results are available for subscription."  But the only reference to a subscription to Morgan Poll I can find goes to a free online newsletter signup unchanged for years, and if Morgan had been releasing results via that list then we would certainly know about that.  Morgan has sometimes offered electorate polling profiles on sale on its site (re which, caveat emptor) but I can't find any method of a pay subscription to the poll.

Then there's "The Morgan Poll has been published this week to restore balance to the political agenda – which is ‘today’ dominated by Newspoll and the call by many for Prime Minister Malcolm Turnbull to resign."  It may not be what it sounds, but this sounds like an opinion polling company openly engaging in the kind of selective release that is a problem with commissioned-poll releases based on private polling (since they only release what it suits them to release, you have no idea if what you do see is representative).  If Morgan has been comparing face-to-face data for years without releasing it and has only just decided to let this one loose for the reasons stated, what did the other ones show? And on what basis is Newspoll-based coverage considered to lack "balance" when - a fairly bad miss in SA notwithstanding - the pollster has such a strong record of accuracy in recent state and federal elections and there is no evidence of it having any systematic house effect?

Then of course there's the gloating about the fact that Newspoll missed the Liberal vote by four points in SA.  Of course, it's true that Newspoll isn't perfect, but the other regularly active pollsters have not shown anything much different (except the Fifty Acres series which obviously had serious problems with its preferencing).  As for Morgan, their last released SA poll was in January.

Finally, it's not clear whether the preferences in this poll were respondent-allocated or last-election, though they are certainly compatible with the latter.  For the time being I have decided to hold off on aggregating this new Morgan series, but I may do so if it turns into a regular free or inexpensive public release.


  1. Hi Kevin, this is OK topic but...

    I'm interested in your take on what happened in Victoria with the broken pair agreement. I thought such an action must surely be unprecedented, but apparently Gorton was involved in something similar. I can't glean much online about that or any other historical incidents. Do you have any insights?

    1. There was one in the Whitlam years that Stephen Murray has mentioned:

      As I understand this one, Carling-Jenkins who was probably opposed to the bill was ill and there is a dispute about whether she asked for a pair. Either she didn't ask (possibly an omission) or she asked and it wasn't granted. The bill was then set to pass because she was absent until the Liberals connived, in an extremely sneaky fashion, to extract a pair from Labor then renege on it. The defeat of the bill is what the will of the parliament would have been had all been present; it would have been tied 20-20. So it seems less outrageous than if a decision was reached as a result that would never have been the will of the parliament had all members been present. Plus I tend to agree that sitting on Good Friday is just silly.

    2. Hi Kevin,

      Again, slightly off topic, but I have seen nothing on this in the MSM.

      The new Tasmanian Lower House has 25 members - 13 LIB, 10 LAB and 2 GRN. Assuming the speaker is from the LIB ranks, that will leave the floor of the house with 12 LIB, 10 LAB and 2 GRN.

      What happens in the event of a tied vote, when LAB and GRN both vote against LIB? Can the speaker vote with the LIBs or does convention say that he has to vote against the motion?

  2. The speakership conventions on tied votes don't apply when a government has a one-seat majority. The Speaker will vote with the government.

  3. Given Turnbull is ahead as preferred PM, does it really make sense for him to be blamed for their TPP losses? Doesn't that polling suggest he's actually more popular than the coalition average, and thus it's the rest of the party that's more to blame, not so much him?

    1. Better PM (often called preferred PM) is a messy indicator that tends to skew to the incumbent PM, whoever they are. If the 2PP is 50-50 then the incumbent PM will normally be about 16 points ahead. For a 2PP of 47-53 the incumbent PM will normally be about level. Turnbull has often pulled above his weight in that regard, but currently isn't doing so by much.

      Turnbull has generally done much better than Abbott on better PM indicators, and this difference applies even once the comparison is adjusted to control for the 2PPs and the personal ratings of the leaders at different times.

      Historically, PM netsat is a better predictor of 2PP than better PM is. Rather mysteriously, it seems to function as a leading predictor (that is, movement in PM ratings tends to happen ahead of 2PP movement). PM netsat is a direct measure of popularity and shows that Turnbull is not popular - his ratings have been consistently pretty bad since the last election, though not quite as bad as Abbott's.


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.