Thursday, September 28, 2017

Poll Roundup: The Clock Strikes Twenty

2PP Aggregate: 53.6 to ALP (+0.2 since last week)
Labor would easily win election held "right now"

This will be a rather brief Poll Roundup by my standards, because apart from same-sex marriage polling (covered in a separate rolling post) there isn't all that much around to see!  We're five weeks on from my previous roundup, and in terms of the prospect of the government recovering before the next election, that's another five weeks down the drain.  Predictively, that doesn't mean a lot, but it is bad news for one particular member of the Coalition: the PM.  He edges another two Newspolls closer to matching the metric of 30 consecutive Newspoll 2PP losses that he used to justify the removal of Tony Abbott.  Just ten to go ...

Precisely what happens if these ten are all lost and the Coalition are still down the tube nobody knows.  Would the whole "thirty Newspolls" thing take on a life of its own in public perception of Turnbull's fate, contributing to even worse Newspolls, or would it only be of interest to the beltway and political junkies, and shrugged off as irrelevant by everyone else?  For it to be game over the very same week, while logical and fair, would seem too obvious, too artificial.  These bad polls seem so set in, and the Galaxy-run Newspoll so remorseless, that it's hard to see just what would end it.  A personal triumph on same-sex marriage? Worth some bounce surely, but enough for 50-50 after such weakness on the issue? War with North Korea? Maybe, though whether the more likely mechanism there is a rally round the flag or Newspoll being hit by an errant missile meant for Guam is not clear either.

Anyway, the latest polls.  In the last five weeks the Coalition has had a 47 and a 46 on two-party preferred from Newspoll, two 50s from the idiosyncratic YouGov, a 47 from the first Ipsos in four months, and a run of 47-47-46-48-47 from Essential.  The YouGov 50s come out as 47.4 and 46 by the nearest I can get to 2016 preferences, and the Ipsos poll and the second Essential were both most likely on the downhill side of 47.  So based on last-election preferences, I now have the Coalition at 46.4% 2PP, not quite their worst for the term, but pretty close.  Here's the smoothed tracking graph:

The Coalition were desperately hoping for a bounce at Budget time, but in fact their best polling of the year came just before the Budget, and the trend since then has been a very gradual worsening.

It might not really be that bad, because of the issue with One Nation preferences (which were historically weak for the Coalition at the last election and may have since been historically strong.)  The Ipsos (which is probably more reliable on the respondent gap than YouGov) had the 2PP at 52-48 to Labor.   It's still bad however you realistically slice it though.  Anecdotally, it's worse: one hears stories of the former Coalition base deserting in droves for One Nation or Australian Conservatives, though no evidence of the latter has shown up yet in polling.

Some reports of a Western Australian ReachTEL have been seen, but it appears the reporting of the primary votes is marred by the usual problem of failure to redistribute the "undecided", and therefore that reports of the primary vote swing against both majors are overstated.  (See figures with undecided excluded here).  The poll isn't as bad for the Coalition as some (perhaps showing some signs of the lifting of the Barnett curse), but they would still be likely to lose a few seats.

Yet more Australia Institute ReachTELs that continue their microscopic study of the renewable-energy pony to no discernable effect have also been seen.  These show small respondent-allocated swings to the Coalition in strongly ALP-held coal seats, which will concern Labor not at all (and even if they did, seat polls are iffy.) A neat summary of the issue findings would be simply that even in coal country, less than half the voters have anything to do with coal and the rest answer like anyone else.  The wording of Question 7 (which asks voters to choose a new renewable energy target, while giving them only the choice of increasing or don't know) was especially cute.


Leadership polling has seen remarkably little movement in recent months - the Bludger Track aggregate shows very little movement for either leader in the past year, and only perhaps a slight widening of Malcolm Turnbull's preferred Prime Minister lead.  Bill Shorten's netsat seems especially stuck in the mud (not for the first time), with three consecutive readings of -20 and seven such readings in the last ten.  The latter is probably some kind of record, but it's really not worth checking.  PM Turnbull's result this Newspoll (-17) was his second best since the election, but hardly a significant improvement.

Turnbull returned a term-high lead as better PM in Newspoll last time around (46-29) but this week it's gone back to a more typical 11-point lead (42-31).  It remains interesting that Turnbull keeps maintaining this lead while his party polls very poorly, but it just doesn't seem to be helping.

Ipsos, which is typically friendly to Turnbull compared to others, recorded a net -5 for the PM (42-47) and a rather nasty -16 for Bill Shorten.  Ipsos was also friendly to Treasurer Scott Morrison, giving him a net rating of +4 (42-38).  Perhaps ALP-lite big government isn't so unpopular after all!  Morrison had a modest 38-29 lead over Chris Bowen (who I haven't noticed all that much lately) as preferred Treasurer.

Essential has somewhat concurred with Ipsos, also giving Turnbull a net -5 and giving Shorten -11.

I may add any other notes of interest to this roundup but in view of time pressures will hit "Publish" now.


  1. While waiting for a new roundup, I saw some dubious poll reporting in The Guardian today.

    Looks like they are turning federal voting intention into a state polling headline. Am I right in assuming that's a big no-no?

  2. Essential does poll state voting intention, which is then reported/released erratically and sometimes several months after being taken. Looks like these are state intention figures.