Tuesday, May 18, 2021

Poll Roundup: Another Budget Rates Pretty Well As Early Election Talk Resumes

With the 2021 Budget now released it's time for this site's annual post-Budget federal polling roundup.  

Voting Intention

The current polling consensus is that Labor has a narrow (51-49ish) lead, though we should not be too confident Labor is ahead at all given the errors seen at the 2019 election.  We can hope that pollsters have successfully addressed the causes of the errors (and also not overcorrected, in which case Labor might in theory be ahead by more). YouGov at least has taken targeted action aimed at fixing the 2019 problems, but until we actually see a test at a federal election, we don't know.

The current Newspoll had Labor ahead 51-49 (unchanged) off primaries of Coalition 41 Labor 36 Greens 12 One Nation 2 others 9.  The second ever Resolve Political Monitor poll (link to my April discussion) has primaries of Coalition 39 Labor 35 Green 12 One Nation 2 Independent 8 Other 5.  Treating "independent" in this poll as generic "others", I get a 2PP of 51.1 to Labor (+1 since April).  If "independent" is carved out, I get 51.9 (+0.9), but I don't recommend this as there are plenty of low-information voters who use "independent" as a catch-all term for indies and minor crossbench parties such as KAP, Centre Alliance etc.   As noted in the initial article about this poll, using Independent as a standalone option can lead to inflated readings compared to actual Independent vote shares.  

The initial Resolve poll had One Nation on 6% and they have now dropped to 2%, with Others up from 2% to 5%.  The April poll was criticised for getting One Nation on such a high figure, possibly by including them in the readout everywhere whether or not they are actually running there.  Whatever the reason, nothing has happened since that would have suddenly cost One Nation two-thirds of their supporters, and low readings are less volatile than high ones (the in-theory margins of error on these results are 1% and 0.7% respectively) so this looks like a methods change.  However I haven't seen any discussion of a possible methods change in the accompanying reporting, which treats it as a change in voting intention: "While Pauline Hanson’s One Nation lost ground, there was a corresponding gain in support for “other” candidates."  Did One Nation actually lose ground, or did the poll handle them differently, or both?

This appears a handy time to mention that the Australian Polling Council, consisting of YouGov (which conducts Newspoll as well as YouGov branded polls), Essential, uComms, Ipsos, Lonergan, Telereach and JWS, will launch a detailed Code of Conduct for pollsters next month that I expect to lead to far more public detail about how Australian polls are running.  At this stage Resolve is not a member of the Australian Polling Council and I very much hope that it will join, or that its client will persuade it to, or failing either that it will at least match whatever disclosure standards the APC releases.

Aside from these two series, there's not that much to see in voting intention land.  Results for Essential were reported in The Guardian on April 29, with "an effective tie between the major parties, with both Labor and the Coalition recording a two-party-preferred vote of 46%," (equivalent to 50-50 based on Essential's 2PP Plus system.)  Other details included were that Labor had led 48-45 (=51.6-48.4) in March and that the Coalition primary had moved between 37-39 with Labor between 34-36.  These would be raw figures without undecided redistributed, so both would actually be a few points higher, as would the Coalition's primary vote among women (quoted as 31 in the April 29 poll).  

However Essential hasn't put these figures up on their website weeks after they were reported, and currently a 47-44 lead to Labor from three and a half months ago is still flapping in the breeze on the sidebar as if it is a current voting intention reading.  

The matter of gender splits in voting has been of interest lately with Essential persistently finding large gaps and Newspoll finding even less than normal.  In the latest Essential sample it's effectively 57-43 one way for one gender and 57-43 the other way for another.  The initial Resolve poll found nothing to see here but the new one implies the Coalition 2PP around 53.6 for male voters and 44.9 for female.  Across the two samples combined Resolve is finding a fairly normal gender gap of a few points.  

One other voting intention finding comes from the quarterly ANUpoll series.  After removing undecided I get Liberal 39.1 Labor 36.8 Greens 15.3 others 8.4, which implies a large Labor lead of about 53.5-46.5, compared to 50.5-49.5 three months ago.  However I don't take polls that have the Greens this high seriously; this is usually a sign of over-sampling of politically engaged voters.  This panel-based survey has very large sample sizes but mostly repeat-samples the same pool of recipients. The polling report has this concerning weightings used:

"1. Compute a base weight for each respondent as the product of two weights:

a. Their enrolment weight, accounting for the initial chances of selection and
subsequent post-stratification to key demographic benchmarks

b. Their response propensity weight, estimated from enrolment information
available for both respondents and non-respondents to the present wave.

2. Adjust the base weights so that they satisfy the latest population benchmarks for
several demographic characteristics."

Good luck to anybody seeking to replicate, or even comprehend, whatever that is!


From its pandemic peak in around mid-2020, Scott Morrison's approval ratings were gradually declining but still very high until March 2021.  From that point there was a crash that strongly appears to be caused by the Higgins/Porter/staffers cluster of issues involving the treatment of women, and now Morrison has recovered to some degree (see the Bludgertrack approval tracker).  He's at net +20 in Newspoll presently and net +15 in Resolve.   The +20 isn't the +41 he was getting mid-last year, but still it's not shabby either.  Morrison's current streak of 18 consecutive positive ratings is surprisingly the fourth longest in Newspoll history (see list here; I don't think he'll be touching John Howard's record of 102).  Meanwhile Anthony Albanese isn't getting a lot of love from the new poll Resolve, which has him at net -12; Newspoll's -7 (his equal worst net rating, albeit for the fourth time) is not much better.  In this week's Newspoll 46% are dissatisfied with Albanese, the highest level so far.  Essential also has Morrison rebounding, from +17 up to +26, while a friendly run for Albanese from it is starting to run out of steam as it only has him on +4 (39-35).  

The skewed Better (Newspoll, Essential) or Preferred (Resolve) Prime Minister questions see Morrison with a current lead in the mid-20s.  Such questions tend to skew to incumbents, and given that Morrison's party is trailing slightly on the 2PP he would normally have a lead of only around 11%.  That he is doing better is partly because of his high personal ratings, while Albanese's personal ratings are mediocre.  


Newspoll asks three Budget questions in a stable form which is great for comparisons between years.

On the question of whether the Budget is good for the economy, Newspoll respondents say yes by a margin of 44-15, for a net +29.  This is the sixth highest on record, behind:

1987 +49 
2007 +48
1996 +37
1998 +35
2006 +31
2021 +29

The top four were election years, but the 1987 budget came after the election.  

On the question of whether the Budget is good for the respondents' own financial situations, it's 19% for better off, 19% for worse off and a record 62% for meh/dunno.  But a net score of zero is actually the eighth highest reading ever.  All the positive scores were for Coalition budgets, including the three before this one.  The only cloud here for the Coalition is that the Budget did include a measure described as a tax cut (though I believe it's actually the extension of a currently existing rebate) so maybe the government would have hoped for a better personal impact score.  On the question of whether the Opposition would have done any better, it's a 46-33 lead for nope, fairly typical by recent standards after last year's Budget recorded the biggest lead for no since 2009.  

The once-off Newspoll questions included one that asked:

"The current budget is in deficit and forecast to remain in deficit into the future, so that net debt will climb from $617bn this year to $920bn in 2024.  The government says the pandemic is not over and this budget will ensure we come back even stronger, securing Australia's recovery.  What is closer to your view about this budget?"

Options were "The government is right to stimulate the economy even though this means increased debt" and "The government should be doing more at this time to rein in spending and reduce debt" 

I don't like this question, but at least it was published verbatim, unlike (so far) a number of the Resolve Budget questions that have been referred to in media reporting.  The question part gives the government's defence of increased debt but no counter-argument from anyone who might oppose it, beyond the size of the debt.  The first reply implies as fact that the government will be successful in stimulating the economy.  So I disregard the 60-30 margin for the first response over the second.  But it is interesting that Labor has at times been flirting with austerity politics in attacking the Coalition over debt (doubtless in frustration that the Coalition got away with the same thing so much more easily around the time of the Global Financial Crisis) and when the question is framed like this, Labor supporters become more likely to be opposing big spending.  

Here's my increasingly scribbled graph showing the relation between economic and personal scores in Newspoll since 1988 (the personal impact question seems not to have been asked in 1987):

Note that the 2007 Budget is the best received of these on both scores, yet the Coalition lost that election.  They did lose by a great deal less than lead-up polling indicated, with only the final polls getting it more or less right (mostly).  

Resolve has asked rare questions about the performance of the Treasurer and Shadow Treasurer, with Josh Frydenberg recording a strong +31 (54-23) net rating while potential leadership contender Jim Chalmers scores an underwhelming -3 (24-27).  Resolve found much stronger approval for the Budget than Newspoll did (56-10 on impact for the country as a whole, 35-17 for personal/household impact) but I haven't seen the exact questions asked by Resolve there.  

The current picture

Talk about, or about avoiding, early elections tends to be very poll-driven.  At the start of the year an early election was all the rage.  During March and April, as the government took a modest polling hit the consensus was that an early election was off the table.   Now that Morrison's own ratings are rebounding, the government's 2PP is competitive and the budget is all spendy-feely, it suddenly seems to be on again.  At the moment the government has good reasons for optimism whenever the election is since oppositions that only lead narrowly don't generally end up winning the election.  

The lack of a "Budget bounce" shouldn't be cause for concern for the government at the moment, since such bounces rarely even exist, and when they do exist, they sometimes take time to become fully apparent (an arguable example being 2019).  Indeed the normal average voting intention change following budgets is a slight decrease, with the exception of Coalition election year budgets, so the next few polls may be of interest.  The biggest potential pothole issue for the Coalition at the moment could be criticism of the pace of Australia's vaccine rollout, though this might also be an incentive for the government to get off to the polls in October or so before the impacts of disparities become too severe.  While polls have showed a mixed reception for the government's vaccine rollout performance (Essential negative, Newspoll slightly positive) the government continues to score well on its handling of COVID overall.  

In the Resolve poll voters say they don't want an early election, but in general I suspect voters only want an early election when they're desperate to throw the government out.  

I may add comments on other federal polling matters should I think of anything else worth noting at this time.  I doubt I'll have time for a preview post about this Saturday's fascinating Upper Hunter state by-election in NSW, but I expect to have live commentary on the night and postcount comments if required.  


  1. My general impression is that the Coalition might gain in NSW, particularly Sydney, but loose ground elsewhere.

    An early election is extremely unlikely to be called before the new financial year, for a dtate no earlier than the 7th of August, due to the Senate election timing requirements in the Constitution (and the legislated requirement for a Saturday election). The redistributions in Victoria and WA are also due to finish on the 26th of July and the 2nd of August, respectively, so any election called before then would involve mini-redistributions. I therefore think it unlikely that an election would be called before the 2nd of August. This leaves a significantly larger budget-election gap than last time.

  2. In the circumstances of the past few months, I wonder if 'Others' contains a higher proportion of Liberal/conservative voters.....either less willing to say they're voting Liberal, or parking their vote for the moment?