2024 Federal Polling Year In Review

 2PP Aggregate Average For 2024: 50.9 To Labor (-3.9)
Labor lost aggregate lead late in the year

It's the time of the year when most busy pollsters take a few weeks off and I bring out an annual feature, a review of the year in federal polling.  Click here for last year's edition and for articles back to 2014 click on the "annual poll review" tab at the bottom of this one.  As usual if any late polls come out I will edit this article to update the relevant numbers.

2024 was another strong year in results terms for the Australian polling industry.  Pollsters came out in good numbers for the early 2024 Tasmanian election and did pretty well in a very hard to poll election, although the lobby group(s) that commissioned two Freshwater polls contemptibly failed to ensure the release of definitive results of either, leaving poll-watchers to play jigsaw puzzles with incomplete media reports. Polling for the Queensland election was mostly excellent though no one pollster nailed the result, and a mini-cluster of close-ish polls at the end led to some misreads of what in the end was not a close election.   Despite this there are a storm clouds about in federal polling in the form of inadequate transparency from several pollsters and a somewhat suspicious level of clustering of results, especially at a stage in the cycle where that doesn't usually happen.  (More on the latter later).  US polling this year wasn't as bad as 2020, but didn't quite get the real story in part because of the latter issue.

There were no major changes in the polling lineup.  But there was a significant change in the behaviour of Resolve relative to other pollsters.  Following the 2022 election (but not in the leadup to said election) Resolve had frequently showed much larger primary vote leads for Labor than other pollsters (not only federally but also often at state level in the eastern states), as a result of which my 2PP conversions for Resolve diverged from my aggregate by an average of 3.7 points from the 2022 election through the end of 2023.  But in 2024 the difference has dropped to 0.4 points for the year as a whole, and 0.06 points for the last nine Resolve polls (effectively nothing).  I would assume some kind of change occurred, but have not seen anything published as to what.  This should be kept in mind when reading Ninefax narratives that talk about how Labor has plunged from (insert very high primary vote number) to their current primary of a miserable 27 - no other polls had them on such huge numbers in the first place, and most other polls don't have them that low now either.  

The current batch

As usual a quick update on the state of play since my last roundup.  After Labor lost the lead in my aggregate in the second week of November, they briefly recaptured it but have now lost it again and currently trail 49.7-50.3 by last-election preferences.  This week's Freshwater (51-49 to Coalition) was little changed from the previous and fairly harmless for Labor by Freshwater standards.  Essential was bad on their respondent-allocated "2PP+" at 48-47 to Coalition (=50.5) but the primaries were fairly similar to the last election after adjusting for the way Essential handles undecided.  The way Essential uses recalled vote creates a high risk of its primaries looking too much like the last election (a common issue with US polling this year) and for this reason my aggregate is currently adjusting its numbers down.  (By last-election preferences the latest Essential came out at 51.8 and nobody else is getting that now).  

The stinker of the week for Labor was the Morgan with the Coalition jumping to a 41-27.5 primary vote lead, the highest primary vote lead of the term for them by a striking 2.5%.  By last-election preferences this came out to a 48.5-51.5 deficit after a run of Morgan being pretty ALP-friendly on this measure.  Morgan is a fairly bouncy poll and not too much should ever be read into its single readings (and perhaps even less so this close to the holidays) but results like this still move the probability dial.  The probability at the moment is that the Coalition is just in front if last-election preferences still hold, and they probably don't given the shift in One Nation preferencing seen in the Queensland election and the Fadden by-election.  My One Nation adjusted figure has had the Coalition nudging 51 lately.  An election held now would not be pretty for the Government - they could well still win, but it would be the kind of victory you don't recover from.

How many polls?

This year again saw a swing to the supply side in federal polling.  Excluding ANUpoll which has been wildly inaccurate and is therefore not included in my tally, there were 132 federal polls (50 Morgans, 24 Essentials, 14 Newspolls, 12 Freshwaters, 11 Resolves, 11 YouGovs, 5 Redbridges, 3 Accent/Research MRPs, one DemosAU and one offering from the enigmatic Wolf+Smith.) That's an increase of 23 compared to 2023 and makes this year the busiest since 2013 (when there were 142). The main causes of the increase were YouGov running separately from Newspoll for the whole year, and Freshwater running monthly polling for the whole year after starting federal polls late in 2023.

2PP Voting Intention

124 of the 132 polls listed above released a headline two-party preferred number of some kind; those not doing so included seven of the 11 Resolves and one of the three Accent/Research MRPs.  Of these 2PPs Labor won 59, tied 28 and lost 37; my last-election estimates for the eight with no released 2PP were six Labor wins and two to Coalition, though one of the Coalition wins would have been rounded to a tie, so let's say +65=29-38 for the Government in chess scoreline terms.  Morgan released last-election preferences as a secondary figure for 26 of its 50 polls; using these instead would have turned five of Labor's losses into wins, four losses to draws and three draws to wins.  Resolve, when 2PPs were reported, varied as to the form of preferences reported and whether one number or two were reported; in one case last-election prefs (a win for Labor) had the most prominence but respondent preferences had Labor behind.  I should note of course here that winning the 2PP is not the same as winning a hypothetical election.  How much it is not the same is a very big question heading into the real deal.  Finally, one of the YouGov 50-50s has been the subject of a comment from a YouGov analyst that the 2PP should really have been 49-51.  This has, however, not been corrected on the YouGov website so I'm continuing to treat it as 50-50.  

Coalition 2PP wins on released figures became commoner through the year with three in the first quarter, six in the second, 12 in the third and 17 in the fourth.  There was some lumpiness by month in this with none in June but seven in July; otherwise November (6) and December (8) unsurprisingly saw the largest numbers.

Many of the Coalition's 2PP wins based on pollster-released estimates would not have been wins by last-election preferences; on that basis the Coalition saluted only 28 times, 18 of which would have rounded to 50-50 to the nearest whole number.  My highest last-election conversion was a 53.9 to Labor for Morgan in early March and the lowest was 48.1 from the September Freshwater offering.  By pollster-published 2PPs the same Morgan topped the list at 53.5 (tied with another Morgan in the second week of June) and the worst was an Essential that converted to 46.8 in late March (thanks to volatile respondent preferences).

My average last-election aggregate figure for 2024 was 50.9 to ALP (down 3.9).  However, again my aggregate liked Labor's old stuff better than their new stuff.  Apart from some weirdness in January caused by a single Morgan outlier at a time when there was not a lot of polling happening, the story of the government's year is down.  From the 52-48ish position it ended 2024 in, the government lost one point between mid-February and early April, hovered around 51 before dropping another half-point in July, and then clung to the tiniest of leads (on my reckoning but not everyone's) until mid-November.  We end the polling year with the government in a third decline phase that has yet to convincingly stabilise. 

It's not too easy to link specific events to the cases where government polling slumped, though the loss of half a point in mid-year did directly follow the departure of Fatima Payman from the Labor Party.  Unlike some other years, there hasn't been an obvious single big poll-shifting movement, and the difference through the year hasn't been that large.  

Herding With No Predator In Sight?

2PP polling has obviously been very close this year, especially in the second half.  But what's odd is how consistently close the headline 2PP readings have been.  I noticed this in early September, and the pattern hasn't changed.  Since Morgan's equal year high 53.5 in the second week of June, the average pollster-released Labor 2PP has been 49.9 +/- 0.9.  There hasn't been a single 2PP outside the range 48-52 in the last 69 readings!  The headline 2PPs have only about 60% of the variation one would expect if they were completely random samples with a margin of error of 3% and if there were no house effects between different pollsters and if there was no underlying change in voter intentions during that time.

If there is some kind of volatility suppression going on (and it could be just from one or two firms if so) the question is why bother.  Pollsters are judged on the accuracy of their final polls and historically that's when herding peaks because nobody wants to be wrong by themselves.  If a pollster produces the most accurate final poll in 2025, surely nobody but the most hardcore statistical purists will care if it threw a 46-54 rogue or bounced like a frog in a sock a few months out?  

One thing I have seen a lot of this year is that lay poll consumers are more likely to notice (and exaggerate the meaning of) relatively minor differences in polls when the polls are tight. I've been getting "the polls are all over the place, I dunno who to believe!" type comments from people on social media when one pollster releases a 51-49 and another the same week releases a 50-50!  (It's quite exhausting actually; I ask those doing this to kindly stop.) I get complaints that Morgan is useless because of its bounciness when even Morgan with an average 2PP shift from week to week of 1.3% is actually not that bouncy (and that average has dropped from 1.58% to 1.04% in the second half of the year, making me wonder if something changed in Morgan's methods at the same time as Morgan started releasing two sets of 2PPs).  

Another is the intensity of groupthink among the commentariat about an inevitable close contest and a nearly inevitable hung parliament, when by historic standards there's still plenty of time for something else to happen.  It may be that any pollsters who might be prone to make a decision influenced by other pollsters, their own prior polls or the zeitgeist would think there is no point putting out a sample in which the race isn't at all close when everyone "knows" it is and will remain so.   

The ALP Primary

One thing that pollsters are not herding on and that has triggered a lot of comment this year is Labor's primary vote.  In the latest reading from each pollster Labor has 27 in Resolve, 27.5 in Morgan, 30 in Essential, YouGov and Freshwater, 32 in DemosAU, 33 in Newspoll and 34 in Redbridge (but 31.6 by my estimate in the Accent/Redbridge MRP).  They can't all be right!

As averages through the year (noting that the time scales are not all exactly the same) Resolve had Labor at 29.5, Essential 30.4 (but really 32.1 after adjusting for "undecided"), Morgan 30.8, Freshwater 30.9, YouGov 31.3, Newspoll 32.6 and Redbridge 33.0 (but Redbridge only from five polls and the latest Redbridge/Accent MRP is lower off a very large sample).  Resolve and Morgan both have "independent" available as an option everywhere, which drags down primaries for other forces (I suspect especially Labor) compared to what happens at an election when voters who might in principle vote independent find there is none or only a silly or obscure one in their seat.  It's possible Resolve's use of forced choice (you can't say "don't know" and continue in the survey) drives up their independent response during the term further.  (Their campaign period polls work differently and tend to be accurate on the IND vote).  So in real terms, rather than the spread of readings of the ALP primary between different polls being the six points it was in the current rount, on average through the year it might be something like two.  Newspoll gets the highest ALP primaries of the monthly or more frequent polls, but not massively so.  

It's also notable here that Newspoll (12.1) has had a lower average Greens primary this year than Resolve (12.5), Essential (13.0 after adjusting for undecided), Morgan (13.3) Freshwater (13.4) and YouGov (13.5).  Among the more frequent polls that don't have high IND breakouts, one can pretty much throw a blanket over the average combined ALP/Green primaries of YouGov (44.8), Newspoll (44.7), Freshwater (44.3) and Essential (43.4 adjusted).  

Leaderships

I usually use Newspoll alone for the annual leadership averages but that comes with a big asterisk this year because Newspoll has in some regards been the odd one out.  In Newspoll, Anthony Albanese averaged a net satisfaction of -9.1 for the year, with a best of net 0 in May and a worst of net -15 in November.  Peter Dutton averaged net -12.7, with a best of net -8 in late July and a worst of net -16 the poll before in late June.  On Better Prime Minister, an indicator that historically favours incumbents by an average of about 15 points, Albanese led every poll, but only beat the historic average once (a 19 point lead in May), with a lowest lead of four points in November; his average was 9.6 ahead.  Albanese in fact has the second-longest Newspoll streak of Better PM wins against the same opponent ever in time terms (behind Turnbull vs Shorten), but that's mainly because Opposition Leaders who were consistently losing Better PM were mostly losing it by lots and getting removed.  

However in Newspoll it took til the final two polls of the year for Dutton to get a better net satisfaction score than Albanese.  In every poll that wasn't a Newspoll since early June (26 in a row from five different pollsters) Dutton's personal rating of whatever form was better than Albanese's - the average of these wins being a far from trivial 11.6%.  The major difference between Newspoll and the others is that Newspoll finds more disapproval of Dutton.  Newspoll's average for the time in question has been 42-53 for Albanese and 39-51 for Dutton, while the other polls have averaged 37-50 and 40-42.  It's possible Newspoll's harder pushing of respondents to answer the question is contributing to more voters expressing mild negatives about Dutton.  Dutton even managed to lead on preferred PM polling in four polls this year (three Resolves and a YouGov) and tie another two Resolves.

Betting

Betting isn't reliably predictive but is interesting to keep an eye on.   The US Presidential election this year was the first example I can ever remember where headline betting and predictive markets turned out to have a better take on things than most of the poll-based modelling (DecisionDeskHQ excluded).  Immediately after the POTUS election the Coalition flipped to favouritism on Australian markets, but I think this could have been more than people just lazily assuming that as goes the US so go we.  Specifically, the aftermath of the US election shone a spotlight on the generally poor performance of incumbent governments in the current global economy.  Checking a bunch of betting sites I found the government's implied chance of re-election down to an average of 43.4%, again with not much variation.  That's pretty amazing for a first-term government that hasn't fallen badly behind in the polls, hasn't spent the term fighting with itself and has the pendulum on its side - especially as if the Coalition falls say four or five seats short of majority, the government will have the option of making its crossbench supporters put their name to it with a no confidence vote on the floor of the House. 

The road ahead

At the moment 2025 looks like a really interesting federal election.  Hard economic times have strongly challenged the assumption that the government will win just because first-term governments do.  The near-constant decline in government polling over more than 18 months belies the fact that the government really isn't far behind and was ahead for a very long time.  (Indeed on my numbers they kept the lead longer than any government in the history of regular polling).  But the fundamentals seem bad - voters strongly think the country is trending in the wrong direction, and Australian standards of living have declined.  In last year's edition I called weak polling near the end of 2023 a wake-up call for the government and raised the question of whether the government could improve and put some heat on Peter Dutton.  They did neither of those things, but nor did they collapse in a heap.  Nonetheless as the year ends things have gradually been getting uglier for them.  

The Government retains three options - an early election in late February or March without parliament resuming, an April election with perhaps a brief resumption or a May election with a full Budget as currently scheduled.  Depending on which path is taken there may not be all that much polling before the campaign kicks off, or there could still be a few solid months of it.  As mentioned in the previous article here, the recent history of close leadup polling suggests that a lopsided result (like say 53-47 either way) would be quite surprising from here, but there's still a good chance one side or the other will have a clear win on the 2PP and (not always the same thing) the seat tally.  As against that, there is still the possibility that neither side will get all that close to a majority.  

The Coalition should be well pleased with how the year has ended, but they should also be nervous about what happens if they start to look like clear favourites.  There's not been that much scrutiny of the Coalition's plans until the recent nuclear debate.  I also think concern is warranted about the weakness of the Coalition's team.  Peter Dutton as leader has done better than many would have expected, but he has had to be everywhere in the media cycle to do it.  His frontbench simply isn't a patch on Labor's and remains a mostly B-grade relic of the Morrison days than a distinctive new direction.  Maybe none of this matters if voters are cross enough with the incumbents and feel that the government has just not done enough to help them through.