Wednesday, April 6, 2022

Poll Roundup: Lukewarm Budget Reception As Coalition Remains Way Behind

Little evidence of "budget bounce"
Labor would comfortably win election "held now" 
Leads in polling at this stage tend to contract

Herewith a roundup of polling following the Budget.  It's been a long time since I had time for the last federal polling roundup here, so I just want to set the polling scene before moving on, time (what is that?) permitting, to Tasmanian House of Reps seat guides, a Tasmanian Senate guide [EDIT: that will come after nominations are released] and hopefully a general assessment of the prospects for the Senate.  In short, the government's current polling is rather bad, but it is not all over yet.  To take things pollster by pollster, with the three who are Australian Polling Council members at the top and the laggards and backsliders at the bottom:


Newspoll returned a 54-46 to Labor, down from 55-45.  The primary votes are Coalition 36 Labor 38 Greens 10 One Nation 3 UAP 3 Others 10.  Off these rounded primaries, the expected last-election 2PP would be 53.5% to Labor, so it's possible that Labor has benefited from the rounding of the 2PP.  It's also possible that, for instance, the Labor primary might really be 38.4, the Coalition's might really be 35.7, and the rounding might be all or mostly on the primaries rather than the 2PP.  (Interestingly the Australian's report bothered to spell out that the 2PP is calculated off the unrounded primaries - not the rounded ones which would be very silly - but I don't know whether that should be taken as a signal that the primaries were better for Labor than the rounded 2PP suggests.)

In the previous Newspoll, Anthony Albanese had recorded a tie with Scott Morrison on the Better Prime Minister indicator, which skews heavily to incumbent PMs (so if an opposition leader is more or less equal, this usually means the government is struggling).  This ended a streak of 32 consecutive leads for Morrison since Albanese had led three times during the far-off early 2020 bushfires.   In this one Morrison was back in front by a single point.   Morrison's personal net satisfaction remains underwater, but at net -12 it's ordinary rather than terrible, and indeed his appeal is quite resilient given the recent economic and COVID turbulence, the range of controversies he's been involved in lately and the collection of insults he's collected from fellow Liberals.  (There is lately some similarity with Donald Trump, whose ratings were usually poor but never as terrible as some past Presidents, no matter how much politics junkies expected otherwise.)  Albanese rates a respectable net -1, and is more popular this year than late last year so far.   For comments on Newspoll Budget polling see below.


The AFR has commissioned a new Ipsos series which is a multi-mode poll, 80% online and 20% telephone (mixed mobiles and landlines).  The first run of the new poll has a 55-45 lead for Labor.  The previous incarnation of Ipsos overestimated the Greens primary at every election it was tested at.  This resulted in a similar 2PP error to other polls in 2019, but in other elections the old poll did not have a noticeable 2PP skew.  The new poll weights for age, gender, location and household income.  At this stage it does not obviously have the Greens primary issue (maybe a little bit but we'll see how it goes).  

Unfortunately the results of the new poll are being reported in raw form, with not only undecided voters kept in the sample but also the 2% who are not enrolled to vote.  This means the primary votes do not compare to those in other polls or past elections, something which enables ever trendy talk about the major party primary vote crashing faster than it actually is.  With undecided redistributed the primary votes are Coalition 34.4, Labor 38.9, Green 11.1, PHON 4.4, UAP 2.2, others 8.9 - pretty similar to Newspoll.


Essential continues to record more competitive numbers for the Coalition than other polls.  Its most recent sample had Labor ahead 50-45 on its unusual "2PP+" metric; this converts to about 52.6% 2PP based on the way other pollsters report.  However on the published primaries (Coalition 37 Labor 36 Green 10 One Nation 4 UAP 3 others 5 undecided 5) I get only 51.9 to Labor by last-election preferences.  I got the previous Essentials at 52.1 then 52.5.  A possible reason for Essential's recent lean to the Coalition compared with other pollsters is its use of a party identification question in weighting.  When a party is on the nose respondents may be more likely to deny having ever identified with the party even if they did, and this is likely to skew results towards the centre.  Since voting intention tends to converge during campaigns anyway, this method could end up looking farsighted if there is the usual narrowing (or could miss the boat if there's a blowout).   Essential's personal ratings are milder than Newspoll's and in its 22 March report it found Morrison on net -3 (up from -5 in February) and Albanese on net +7 (up from +3).  Morrison is yet to lose the Better Prime Minister lead with Essential, but it is shrinking (most recently to 3).  

I add that while some pollsters (see below) have estimates of independent and others votes that seem too high, Essential's seem too low.  "Others" defined to exclude Greens, One Nation and UAP polled 8.31% in 2019 and there is clearly a concerted and well-funded attempt to build the Independent vote.  While the number of micro-parties running will be reduced slightly following the recent party registration changes and other deregistrations, it would be surprising if the collective Others vote went down, yet Essential is generally finding it at 4-5%, now and then 6%.  


Resolve Political Monitor (not a Polling Council member) did not release a poll in March.  Its April post-Budget poll came out with the Coalition on 34, Labor 38, Greens 11, One Nation 2, UAP 3, IND 9, others 3.  Resolve has very high Independent readings, probably because it forces respondents to give a voting intention and does not allow a don't know/informal option.  I estimate 55.6 to Labor, which seems rather high given that last year Resolve was often more Coalition-friendly than other pollsters (mainly by having very low Labor primaries).  It could be that those picking "independent" are more left wing than those who did in 2019, in which case the 2PP for Labor might be higher.  My estimate for the February sample was 53.9.

Resolve this week had Albanese with a one-point preferred Prime Minister lead, the subject of a very silly front page that declared "Labor hits the front".  No, Albanese had hit the front on an indicator skewed against him.  Labor has led even on primary votes (let alone estimated 2PP) in Resolve since December, and its last three polls have been landslide territory if repeated on polling day.  This is another case where Resolve's failure to publish a 2PP (out of a supposed aversion to horserace commentary) simply leads to the SMH/Age persisting with horserace commentary but doing it worse.  On personal approvals, Morrison is net -14 for performance "in recent weeks" (though I'm not sure how much respondents focus on the recent weeks bit), while Albanese is net -4.


Morgan's (not an APC member either) federal voting intention poll releases have been remarkably abundant lately, with a massive seven of them seeing the light of day since mid-February.  This included a before-and-after experiment re Russia's invasion of Ukraine which unsurprisingly found no evidence of impact.  Major world events that clearly impact Australia often produce polling rallies for governments - eg Tampa, 9/11, global financial crisis, Bali bombing but in this case the immediate impact on Australia would have been elusive.

The most recent Morgan had a respondent-allocated 57-43 to Labor, up from 55.5 pre-Budget, with Coalition 33 Labor 39.5 Green 11 One Nation 3.5 UAP 1 IND 9 others 3.  Morgan polls tend to skew to Labor and have high Independent and Green votes.

Budget bounce?

I find the average voting intention according to these five pollsters, and using last-election preferences rather than respondent, to be 54.6 to Labor.  That doesn't mean that necessarily is what voting intention is out there in voter-land, but it is probably somewhere near that, and the variety of estimates from different pollsters is a good thing.  The business of house effects (estimating how much particular polls might skew to one side or the other) is very tricky now because all the polls have made substantial methods changes since the 2019 failure.  If we totally ignore those changes and also changes of ownership, just considering the brand names that are active right now it might be expected that the current numbers overestimate Labor by about 1.4%, meaning that in an election held right now Labor would only win about 83 seats instead of 89.  However Essential is clearly behaving very differently at present to at previous elections, and the only major continuity between the pre-2016 Newspoll and the current one is the questions asked.  It's also possible that pollsters will react to what happened in 2019 by overcorrecting, though at the moment I don't think this is happening.  

Is there a Budget bounce?  Newspoll and Essential have had slight movements to the government on 2PP, while Resolve and Morgan have had movement away.  Newspoll and Essential are entitled to a higher weighting on past form, but the movements in the others are larger, though both are volatile polls and the last Resolve is a long time ago now.  Overall there isn't clear evidence of a Budget bounce at this stage.  The general pattern with Budget bounces is that most claimed sightings prove to be illusions, Governments do not on average get any dividend from Budgets, but (and it's a very big but) most cases where bounces seem to have occurred have been in Coalition election years.   

The Budget itself has polled lukewarmly.  On Newspoll's question about voter impressions of how the Budget would impact them personally, it scored a good net +1 rating, making it five years in a row the Coalition has rated net zero or better on this matter (not bad going when the only previous such cases were in 2004-2007).  However on impressions of economic impact it scored a lukewarm net +10.  On the question of whether the Opposition would have done better, 40% said yes and 42% said no.  Since the latter would include voters who thought the two parties would have been the same, this is a woeful result for the Coalition, the third worst after 1993 and 2013 and a strong sign that the government is struggling badly in an area where the Coalition should dominate.  See Poll Bludger graphs and comments here .  Here's my much scribbled graph updated to show how the Budget fared on personal and economic impact scores:

So is this government just toast?

The short answer is not yet.  While Labor appears to be in a much stronger position now than at the same point in 2019 (where I had its position as 52.5% by 2016 preferences, or 51.9% after accounting for One Nation preference shifting) sharp voting intention changes can happen during election campaigns, and they usually see whoever is trailing make up ground.  On average a party that is ahead might be expected to lose about two-thirds of its 2PP lead, but that comes with a lot of variation, including cases like 2019 where the party that was ahead ended up losing, and cases like 2013 where the leading party increased its lead.  (The latter was because the Rudd-return bounce was wearing off as Labor's lack of campaign fitness became apparent.)

The government didn't much improve its national polling position in the 2019 campaign - most of its recovery was down to polling error, and we'll never really know but it may have really been winning all through the campaign.  

In terms of historic recoveries, nobody has polled a 46% 2PP this close to an election and won in Newspoll history.  This is the fifth Newspoll in a row for which the 2PP has been worse than one any government has won from at that stage.  But just because records have to be broken to win, doesn't mean they can't be.  Here are some examples showing that recovery could be possible:

* The Hawke government 1987 led 58-42 four weeks out and 56-44 two weeks out in 1987 (estimated 2PPs) but only won the 2PP 50.8-49.2.  However, the old Newspoll might not have been very good back then.  (It was to later get much better.)

* The Keating government 1993 trailed 46.5-53.5 four weeks out and won the 2PP 51.4-48.6.

* The Howard government 2001 led 57-43 seven weeks out, 56.5-43.5 five weeks out and 55.5-44.5 four weeks out and only won the 2PP 51-49.  (This was the bounce for the S11 attacks going away).

* The Howard government 2004 trailed 46-54 eight weeks out and won the 2PP 52.7-47.3.  However the 46-54 was by respondent preferences (really more like 47-53) and also it was a rogue reading.

* The Howard government 2007 trailed 42-58 five weeks out and lost the 2PP 47.3-52.7.

* The Gillard government 2010 led 55-45 five weeks out and only won the 2PP 50.1-49.9.

As an aggregated polling position, 45.4 is worse than the positions of all the Newspoll era governments that won, but also worse than two of the three that lost.  (Howard in 2007 was perhaps slightly worse).  Excepting 2001, recent elections haven't seen shifts of more than a few points in the last seven weeks, but that's not to say it couldn't happen.  And even the same shift from polling at this point to final outcome as seen in 2019 would see the Coalition lose the 2PP very narrowly and maybe just hang on.  

However, to say the government could recover is not to say that it will.  Even in the final weeks very few federal elections reach the stage where the winner is more or less certain.  A recovery from this point - probably involving some combination of improved polling, polling error and getting lucky on the 2PP to seat conversion, would be around the high end of the range.

Weirdly, despite the Opposition being about as far ahead as it could credibly be expected to get (further ahead than Howard's Coalition was of Keating's Government around the same time), one cannot turn over a rock without finding three commentators and a C or F grade pollster who are convinced the election not merely may but will result in a hung parliament.  It could yet do so, but for all the elections that see hung parliament speculation, very few deliver (with another emphatic non-delivery in South Australia a few weeks ago).  The 2PP range a result has to land in for a hung parliament to occur is pretty narrow, and even landing in that range doesn't ensure one.  

We still don't know if the polls have been effectively repaired.  They are more transparent and more diverse, but it's always possible that the difficulty of reaching certain kinds of voters is insurmountable and those voters will break heavily to the Coalition.  I don't think so though - not so much because of the evidence at state level (where accurate state polls can be followed by bad federal polls and vice versa) - but rather because if there are such breaks in the voting intentions of unreachable voters, this doesn't look like being the campaign to bring them out, in the way the last one was.  

In some ways the government has a great story to tell.  Two years ago mass unemployment was widely being predicted; the government avoided that and now the unemployment rate is low.  The government was also successful enough in managing the early stage of the pandemic that Australia's death rate from COVID-19 ranks among the lowest among comparable nations, around a third of the world average.  However, the good economic stats aren't translating into wage rises or relief from surging prices (especially of fuel).  Perhaps the government can turn around pessimism on that front in the final weeks but Essential's budget issue questions find that their respondents think Labor would do a better job of controlling cost of living, cost of housing and unemployment, and also encouraging wage growth.  It's also not promising for the government that its current bad polling isn't just a flare-up but rather a continuation of a trend against it that had been growing since last year and still hasn't seriously turned around.  

The global security situation is another wild card and one of the questions I am sometimes asked is whether World War 3 will save the government.  (This optimistically assumes there would be voters left to save it.) There's an assumption that should events surrounding the Ukraine invasion take a still darker turn quickly that the Coalition will be favoured, but it's worthwhile looking at previous world wars.  The Cook government called an early election just before the outbreak of World War 1, was attacked for unpreparedness during the campaign, and lost.  The first Menzies government went to an election in 1940 and lost its majority (which later led to its demise).  Khaki elections only advantage those governments who can use them to their strength.  

Seat betting

I don't have time for as detailed a coverage of seat betting as last time and won't be updating colour coded charts on it during the campaign but I will make some brief notes so that after the election I can again look at how it performed.   My reviews of the last three elections have all found that seat betting displayed no special insight (indeed in 2013 its final predictions were much worse than most other things doing the rounds.)  At the moment across a couple of betting markets (Sportsbet and TAB) this is what is being expected.  A "close" seat is one with two parties at 3.00 or less on some exchange, or at least one exchange with no party under 1.50.  A tied seat is either tied on all markets or has both parties favourite somewhere.  I treat Hawke as Labor and Hughes as Coalition.  

Close expected ALP gains (9): Reid, Robertson, Longman, Boothby, Chisholm, Bass, Braddon, Swan, Pearce

Tied (1): Deakin

Close expected Coalition holds vs ALP  (12): Hasluck (tied in one market), Tangney, Banks, Sturt, Bennelong, Lindsay, Brisbane, Dickson, Flynn, Casey. La Trobe, Bonner

Close expected Coalition holds vs IND (7): Flinders (tied in one market), Curtin, Nicholls, Goldstein, Kooyong, North Sydney, Wentworth

Close expected Labor holds vs Coalition (3): Dobell, Gilmore, Corangamite

So based on favourite status there are only just enough Labor gains projected to put them over a majority (78.5 seats expected).  Adding in close seats makes little difference to that, but also projects the Coalition to drop a few to indies, for a parliament around 79-64-8.  This is consistent with headline odds that claim Labor has about a 70% chance of winning, because if a party isn't a very clear favourite then it's most likely if it does win then the margin will be modest.

Seat polling

A brace of seat polls published in the Murdoch tabloids (page 1, page 2) came from a robopollster called Telereach trading as KJC Research.  The same pollster has also been known as Mediareach and is little tested at elections.  A number of seats were described in the reporting as "too close to call" but on my estimates off the primary votes the polls had the government losing Boothby, Reid and Swan, holding Bass, Chisholm and Flynn and failing to recover Dunkley and Gilmore.  The implied swings were much more variable than I think they will be in reality.  

Australia Institute uComms polls of Sturt and Boothby were reported (52-48 and 57-43 to Labor respectively) and it's pleasing to see TAI has joined the Australian Polling Council and will presumably now improve its transparency in publishing results.  The robopollster is coming off an utter shocker in the SA seat of Stuart where it underestimated the vote for independent Geoff Brock by the small matter of 36.5% after reallocating undecided.  This may have been partly because Stuart was redistributed, and also the poll was taken several weeks prior to the election, but it is hard to believe anything other than geographically unrepresentative sampling within the seat would have done it (and even then not without some help.)  See the "regional variation" section of my article about why seat polling is unreliable for an insight into possible problems here.

Climate200, in a desperate bid to annoy me into voting for the most coal-burning candidates I can find, reported that their Redbridge polling has been finding Josh Frydenberg in Kooyong below 40% but would not release primaries for any other candidate.  The methods statement released by Redbridge showed that independent Monique Ryan was named in readout but other candidates were not.  In my view this is a defective poll design - leading independents should be named, but if naming them, name other leading candidates as well, especially very high profile candidates who will pull in voters by name recognition that would not go automatically to their party.   Redbridge writes:

"Given the extremely high profile nature of the local Independent, Monique Ryan, the survey was designed to ensure an accurate and sound collection of responses based on this seat’s current political tempreture [sic]."

And they did not think to at least name the Liberal candidate in case some voters had heard of him too???  He's only the Treasurer and Deputy Liberal leader; he can hardly compete for name recognition with a mere local independent.  

As I release this page there are a few more: Australia Institute uComms of Braddon with Labor ahead 53-47 on respondent preferences, but my last-election estimate is 50-50; primaries are Liberal 35.8 Labor 34 Green 5.5 PHON 7.3 UAP 3.1 JLN 7.9 Garland 2.6 others 4.6 with unallocated redistributed.  Also have seen the West Australian has an Utting Research (not an APC member and so this is about all we will ever know) poll of Curtin with Celia Hammond ahead only 51-49 over independent challenger Kate Cheney (primaries 42-24 with 20 for Labor, 9 Greens, 2 UAP 3 not stated).


  1. Would welcome any commentary you may have re the Senate

    1. That's coming - separate article perhaps as soon as tomorrow.