Sunday, May 19, 2024

Budget Week Rolling Poll Roundup

2PP Aggregate 51.2-48.8 to ALP (last election preferences)
Pre-Budget aggregate was 51.0-49.0 
(Topline number for this article frozen as of 28 May, Budget week now being well and truly over.)

Note: False claims have been published by The Australian, Sky News and others about Newspoll, see the Newspoll section below.  

I haven't paid a lot of attention on this site to federal polling in the first half of this year, but with the Tasmanian election out of the way Budget week presents an opportunity to do that.  This will be a rolling article where as polls come out through Budget week and through the current polling cycle I'll track whether anything happens.  Media speculation about a "budget bounce" seems to have died down a bit over recent years, perhaps because evidence has sunk in that it rarely happens (and when it does, it's usually in a Coalition election year).

Nonetheless, some pre-Budget commentary was seeing the Budget through the lens of a potential early election in the second half of this year.  Some anonymous Labor sources are supposedly drawing confidence from the Coalition's failure to get into the lead this year, and starting to think that if they strike at the right time they can win another majority.  

If the Budget polls are poor for Labor that could well go back into its box very quickly, but it's worth a quick recap of the early election scenarios.  The current wave of redistributions is not expected to complete until mid-October, which in theory could be the jump for a late November election, but there is also the complication of an interest rates decision in early November.  There has been some speculation that that could be the jump-off (if favourable) but that would mean December 7 unless the writs were issued by November 4, the first day of the two-day Reserve Bank meeting, or otherwise December 14 (which I'd expect is untenable).   

If the government wants to go really early then that will trigger a mini-redistribution (a rather crude method of reassigning seats to states without the normal full process). Another reason not to go very early is that in Queensland, and perhaps depending on timing NT, voters will not have yet taken out their frustrations on their state/territory governments.  Incidentally if the mini-redistribution did happen, under existing legislation, teals would lose one notional seat through a temporary merger of Warringah and Wentworth, Labor would lose one through the temporary merger of Chisholm and Higgins, but Labor could gain at least one through the splitting of Hasluck and Durack into three.  (Southern Durack is very pro-Liberal, so joining a third of Durack and a third of Hasluck could cause either the merged seat or the remainder of Durack to become notionally Labor, or neither).  

(See Antony Green for the case for why an early election is really stupid.  I'm not so sure that calling an election for late Nov straight after the new boundaries are final is necessarily silly, since the broad shape of new boundaries tends to be well known anyway, but I take the point about Australians being too obsessed with that primitive electoral crud between Mexico and Canada to want an election called in late October.)

Polls below are listed with the most recently seen at the top

Morgan Second Week

The Morgan released 27 May is no longer really in the post-Budget cycle but it is worth noting for its very low ALP primary of 28.5 (this said Morgan tends to get weaker major party primaries than some of the others), compared to 37 for the Coalition.  The rest are Greens 15 ON 6 IND 9 others 4.5.  This is the worst Labor primary from any poll in the term, and Morgan's respondent 2PP came out at 51.5 to Coalition (the Coalition's second best of the term after an Essential).  However by last-election preferences the high Greens and IND votes mean this is still 51.2 to Labor (50.3 if IND/Others are lumped).  The impact on my last-election aggregate is nothing, but only just: a better Morgan from a fortnight ago dropped out of sample meaning that the before-rounding estimate dropped from 51.24 to 51.16, both of which round to 51.2.

Redbridge MRP

The last poll out in the post-Budget flurry is a Redbridge MRP with a large sample size of around 4000.  An MRP (also used by YouGov at the 2022 federal election with a much larger sample size, and by FocalData for the Voice) uses small samples from individual seats and networks these samples with other seats with similar characteristics to form a projection for every seat.  Some of the individual projections will always be rubbish (the system struggles especially with local non-classic contests, though this Redbridge effort seems to have a decent feel for teal seats and rural indie seats - the issue is not being able to anticipate the impact of locally focused campaigns).  

Overall the poll has 52-48 to Labor off very familiar looking primaries of ALP 32 L-NP 36 Greens 13 others 19.  Redbridge has so far been slightly more favourable to Labor than the average but not by much (I get an average of 0.8%, with 0.6% for the last six polls).  The seat projection is for 77 seats and what is said to be a roughly even chance of an ALP majority - most notably the poll gets very little swing in WA which is at odds with general expectation; it also gets relatively little swing on various other aspects canvassed.  On the primary numbers given Brisbane would actually be Labor vs LNP not Labor vs Green, and I'd expect Labor to win easily in either case - that said it looks like the model tends to flatten out the Green vote with some poor votes in good seats and increased votes in weaker seats, something that has been a quirk of Redbridge breakdowns before.  I also doubt an independent would win Cowper based on the published primaries, but perhaps they would in the unlikely case that the Others vote was actually extremely concentrated with them.  This poll is a snapshot under current boundaries and Redbridge say they'll do another one for the new boundaries.  Unsurprisingly the current MRP says that an election "held now" would give a zero chance of a Coalition government.  

I initially included this poll in my aggregate but have taken it out because of the spread of sampling dates (February through to May).


Essential had a 47-46 "2PP+" lead for the Coalition (unchanged) but this was again off a rather stronger flow of respondent prefs than I would consider likely (the primaries of ALP 31 L-NP 34 Green 10 ON 8 UAP 1 Ind/other 8 with 6 undecided convert to 51.3 to Labor by last-election prefs).  I haven't updated my aggregate spreadsheet yet but that it will be either 51.2 or 51.3 when I do (edit: still 51.2).

Essential respondents were pessimistic (30-59 pre-budget and 27-64 post-budget) about the Budget making "a meaningful difference on cost of living" (a question that seems likely to elicit a pessimistic answer) but approved of all the individual measures and also (57-35) said that cutting the migration intake was likely to make housing more affordable.  What that question doesn't mention is that the cut was from an inflated post-COVID high that was not going to be sustained anyway.  

YouGov Q+A Poll

Something a little different, a YouGov poll commissioned by ABC Q+A (which seems a little unusual but better than them using horrible opt-in social media "polls").  Of interest here is the question where respondents were asked to write anything they had heard of in the Budget that would benefit them and 52% were able to name something, most commonly power bill rebates followed by tax cuts.  I can confirm that respondents here answered the question in a blank box; there were not a list of items to tick.


After four consecutive 52-48s, this week's Morgan came in at only 50.5-49.5 for Labor.  But Morgan uses respondent preferencing; my last-election estimates for these five polls were 53.1, 52.5, 51.0, 52.2 and now 52.0.  As the 51.0 drops out of my aggregate this week, Morgan actually results in another slight improvement for Labor on my estimate (now 51.2).  Still, not exactly much of a bounce.  Primaries were 30.5 for Labor, 37 L-NP, 14.5 Green. 5.5 ON, 8.5 IND and 4 Others.  


A 50-50 from Freshwater (unchanged) off primaries of ALP 32 Coalition 40 Greens 14 others 14, but my aggregate thinks this is actually not bad for Labor.  Firstly the primaries are better for Labor than the previous poll with the same 2PP (such that my average last-election estimate for these primaries is 50.5 compared to 49.6 in April) and secondly as Freshwater has tended to run about a point lower than the aggregate, it adds a point to compensate.  As a result my aggregate goes up to 51.1.  Phil Coorey of the AFR has confirmed the 2PP in this case was close to 50.5 before rounding.  

Freshwater gets a net positive of +1 for perceived effect on  household finances (24-23), but joins Newspoll in finding voters are far more likely to think the budget will worsen inflation (39-11).  Freshwater concurs with Newspoll in having an increased lead for Albanese as Preferred PM (but only a modest lead in this case, 46-37) and has Albanese and Dutton both at net -9.  

(PS 27/5 Freshwater have also added an excellent Data Stream page.)


Newspoll has come in strongly for Labor at 52-48 with primaries of ALP 34 Coalition 37 Greens 13 and I'm still waiting for One Nation's.  Also, Anthony Albanese recorded his first non-negative netsat since October, returning to net zero (47-47) while Peter Dutton is at net -12 (38-50).  Albanese has a pretty strong 52-33 lead over Dutton as Better Prime Minister, though most of that is incumbency advantage and the lead is not that high given the 2PP.  

The budget figures are a mixed bag with a rather low net zero rating on the Budget being good for the economy (27-27) but net -2 on personal impact (27-29), the latter being the best net personal impact rating for a Labor budget ever.  However, the Coalition did much better on the opposition-better question than in the previous two (Yes 37 No 43), though not as well as Labor did before the 2022 election.  

Please excuse the much-scribbled graph (I will redo it from scratch one of these decades) but here's the relationship between net personal and economic scores for every Budget Newspoll that has asked both questions.  The difference between net economic and net personal scores (2 points) is the lowest ever, so the message seems to be that voters are noticing the goodies (or at least the absence of baddies) while pessimistic about whether Labor is fixing the economy.  By ALP standards, this was close into getting into the top-right quadrant, but not quite!

Following the up-down nature of my aggregate lately, Newspoll cancels out the 0.6 point drop caused by Resolve and puts my aggregate back to 51.0 to ALP.  

A widespread claim (started by Simon Benson) that 27% of Newspoll respondents is the lowest proportion ever finding a Budget to be good for the economy is FALSE.  Benson incorrectly claims that Newspoll started measuring Budget reaction in 1999; it was actually 1986.  The latter false claim has been seen before and seems to result simply from one of the Australian's previous tables starting in 1999.  While 27% is very low, lower figures on the same question were recorded in 1991 (22%), 1992 (25%), and 1993 (20%).  What is actually a record is the 46% for the Budget being neither good nor bad for the economy, that's the highest ever.  


The first post-Budget poll out was Resolve with ALP 29 Coalition 36 Greens 12 One Nation 7 UAP 2 Independent 12 others 2.  (As always with Resolve, Independent is likely to be inflated compared to what would happen at an election because many seats do not have prominent independent candidates, and it's also possible that Resolve's forced-choice method contributes to a higher Independent vote although election leadup evidence so far hasn't shown that.)  My last-election estimate is 51.1 to Labor, down 1.0, from a poll that so far this year has been still fairly favourable to Labor (see below.)  While my house effect adjustments based on Resolve's past form are waning as Resolve comes back to the pack, they were still enough to knock my aggregate down to 50.4 for Labor.  (Mark the Ballot has discussed an alternative approach of simply treating 2024-model Resolve as a new poll.)  Leadership numbers were unflattering to Anthony Albanese whose net rating for performance "in recent weeks" was down from -2 to -10, with Peter Dutton down from -2 to -3, and Albanese's lead as better PM (which favours incumbents) shrinking slightly from 41-32 to 40-32.

Resolve found favourable results on whether respondents thought the Budget was "good for me and my household" (40-21), good "for the country as a whole" (41-21), good "for the health of the economy" (38-21) and good "for fighting inflation" (34-27).  

Resolve also put out some immigration polling which I found to not be very useful - indeed I generally think the ratio of noise to signal in polling on this subject is very high.  When presented with statements about migration levels respondents tend to think the current level is too high, and when after that presented with statements about projected levels 50% still think the projected levels are too high.  But this usefully doesn't sample what Australians generally currently think about the issue, since most Australians would not know any of these figures.  It only models what would be their immediate response if presented with a certain summary of the information - but in practice voters will receive a range of summaries, different summaries or none at all amid a sea of other political claims. (This isn't just this poll or this issue = I am critical of the majority of media coverage of the majority of specific-issue polling for the same reasons).  

Note: there's a subjective debate about how to do last-election preferences with polls that ask for Independent and Others separately but on the ballot everywhere, as Morgan and Resolve do.  I treat "independent" as independent; treating them both as combined IND/others gives estimates that tend to be about a point worse for Labor.

The "before" picture

This is a summary of the state of polling before the Budget polls started to drop, mostly written the afternoon of the traditional post-Budget Newspoll.  

The Albanese government enjoyed a long post-election honeymoon that ended around last September and was followed by a narrowing in the two-party picture, finishing last year with an estimated 51.7-48.3 lead by last-election preferences.  In my aggregate, what has happened this year so far is close enough to nothing.  There was a narrowing over summer caused by a single bad Morgan poll at a time when no other polls were in the field, and apart from that there was a slight further narrowing from high 51s/low 52s to low 51s in late March.  In the last eight weeks the government has had some moderately good polls and some moderately bad ones, but they have tended to cancel out.   By last-election preferences the government has been as low as 49.4 and as high as 53.1 in this time (or 53.6 if one includes Resolve) but any sorta-bad poll tends to be followed soon enough by a pretty good one and vice versa, so my aggregate has tended to stay around the low 51s, give or take half a point.  

There are some points to note here.  Firstly the difference in primary vote gaps and expected 2PP between Resolve and the other polls has been far less pronounced this year than it was between the 2022 election and the end of 2023.  I don't know if there has been any substantial methods change at Resolve (none has been documented); whatever the reason the three pre-Budget Resolves this year have had an average last-election-expected 2PP difference from my aggregate of 1.5%, compared to an average of 3.7% in 2024.  

Another significant story continues to be last-election vs respondent-preference differences.  If voters for given minor parties or independents will actually give preferences to Labor more weakly than voters for those parties in previous elections, then last-election methods may be overestimating Labor's position.  However, respondent preferences don't work very well at estimating the shift when it does happen, and often predict shifts that don't occur.  This year so far, Morgan's respondent preferences have run a very plausible sounding 0.6% below the last-election estimate (compared to 1.0% between the Voice defeat and the end of 2023), while Essential's have run 1.8% below (compared to 1.3%).   

I'm mildly sceptical of Essential's recent respondent preferences.  For instance their overall preference flow to Labor in the last four polls is 55%, down from 61.6% at the election, and this even though (i) Labor's flow has always been above 60% since 2004 except for falling just below in 2019 (ii) Essential's minor party primary vote split is such that all else being equal the flow would be stronger not weaker.  (A note that Essential's model is not purely respondent-preference based; respondents who are unsure of their 2PP choice are allocated per the last election).  I suppose it's just possible that a very strong flow to the Coalition off One Nation together with some cooling from Greens and independents could do the trick but that seems unlikely - it would be interesting to see what splits Essential are getting from specific parties.  

Resolve also released a respondent preference 2PP for the April poll (50-50) but this seems to have been a once-off.  Although that implies a similar difference to Essential it's hard to know whether it is typical or not, especially if the decision to include it was a once-off post hoc decision.

No comments:

Post a Comment

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.