Until recently, pollsters in Australia haven’t had a lot to complain about. True, sampling isn’t what it used to be. When polling was conducted face to face, and sampling points drawn from the electoral roll gave almost every voter an equal chance of being interviewed — indeed, even after polling drew on lists of telephone numbers — samples were much more likely to be representative than they have become in the age of the internet. Now, pollsters are dependent on samples drawn from relatively small lists of potential respondents, interviews are mostly conduct online, and response rates are very low. Samples may be assembled using quotas, but responses are still so skewed — too many old voters, too many university graduates, too few able to manage written English (in a country where less than half the adult population is functionally literate) — that the data need, where possible, to be heavily weighted.
But what pollsters have going for them should make them an object of envy to their counterparts elsewhere in the world. Among Australians eligible to vote, enrolments are rising: between 2013 and 2025, the figure rose from 92.4 to 98.2 per cent. Compulsory voting largely eliminates the need to work out which respondents are enrolled to vote and will turn out, and which aren’t enrolled or won’t vote. And while pollsters need to report first preferences for a relatively large list of entities — Labor, the Liberals and Nationals (preferably combined), the Greens, One Nation, and others — they are ultimately judged by their ability to estimate just one figure, the two-party-preferred vote, a figure that has proved easier to get right than the first preferences.
Now, some things are changing, and not in ways that make the pollsters’ work any easier. In 2025, turnout (90.7 per cent) was slightly lower than in 2010 (93.2 per cent); of those eligible to vote, less than 90 per cent did so. And 5.6 per cent of the votes cast were deemed informal (something few polls, if any, attempt to measure), slightly higher than in 2022. All up, a formal vote was cast by just 84.1 per cent of eligible adults. Without the Australian Electoral Commission’s efforts to boost enrolment, this figure might have dipped below 80 per cent.
Surveying the pieces or assuming nothing has changed?
Worse was the challenge posed by the Coalition’s disintegration — first, for eight days in May 2025 and then from 22 January to 8 February this year — before coming together again, however shakily initially and desperately now. Should pollsters have carried on estimating the vote for the Liberals, Nationals and the Queensland Liberal Nationals jointly — or should they have begun estimating support for the various members of the (former) Coalition separately?
Perhaps assuming the Coalition would sooner or later be back together — or perhaps because it was much easier — some pollsters continued to report the “Coalition” vote. (Almost no polling was done during the first split.) They either reported a single figure for the Coalition (Essential) or separate figures for the Liberals, the Nationals and the Queensland Liberal Nationals (RedBridge, YouGov).
Two pollsters did reject the business-as-usual model and reduced the Coalition to two parties, Liberal and National. One was Pyxis, which is responsible for Newspoll, the most watched of all the polls. Between 5–8 February it measured what it said was support for the Liberals and support for the Nationals and reported a historic low: just 18 per cent combined, down from 31.8 per cent at the 2025 election. The other was Morgan, which reported support for the Liberals plus support for the Nationals dropping first to 22.5 per cent (19–25 January), then to 20.5 per cent (26 January–1 February).
But measuring support for the parties that had made up the Coalition (even leaving aside the Northern Territory’s Country Liberal Party, or CLP) is less straightforward than this implies.
The biggest problem is that none of Queensland’s seats is contested by the Liberal Party or the National Party: they are contested by the Liberal National Party, a separate entity that styles itself as a “branch” of the Liberals and an “affiliate” of the Nationals. When the Coalition split, the LNP (Qld) stayed together. Of Queensland’s thirty seats, the LNP holds sixteen — ten occupied by MPs who sit in the Liberal party room, six by those who sit in the National party room. Were the LNP to win Kennedy, now held by Katter’s Australian Party but originally won by Bob Katter on behalf of the National Party, the new member would join the Nationals party room. Were the LNP to win any of the seats held by Labor (twelve) or the Greens (one), the new member(s) would join the Liberal party room.
Asking Queenslanders how they would vote, Newspoll divided seats into two categories. In the ten “Liberal” seats and all the other “non-National” electorates the party contested in 2025, respondents were given the option of the Liberal Party (alongside Labor, the Greens, Pauline Hanson’s One Nation, Other parties or independent candidate, and Don’t know); in seats the “Nationals” held and in Kennedy, it offered the option of the Nationals (plus Labor, the Greens, Pauline Hanson’s One Nation, Other parties or independent candidate, and Don’t know). Morgan did something similar, except it offered both alternatives — Liberal and National — in every Queensland seat.
What respondents made of being offered a Liberal and/or National option but not the LNP, the Queensland party that has contested every election since 2010, can only be wondered. Even in seats held by the LNP (Qld), how many voters would know that their member was “really” a Liberal or “really” a National?
Who’s most at risk from One Nation?
The Nationals are widely regarded as the party most likely to lose votes — and seats — to One Nation; and that means the level of support for the Nationals matters. At the 2025 election, contesting twenty seats outside Queensland, Nationals candidates won 3.8 per cent of the vote; contesting seven seats within Queensland, including the seat held by Katter, those who sat (or aspired to sit) in the Nationals party room won another 1.8 per cent. Adding these two figures, the Nationals might claim to have won 5.6 per cent of the 2025 vote.
To what extent has the rise of One Nation been at the Nationals’ expense? Between 22 January and 8 February, the polls estimated support for the Nationals (outside Queensland) to be just 1 per cent (YouGov) or no higher than 2 per cent (RedBridge). On these figures, support for the Nationals had (at best) almost halved since the election. If we include the Queensland figure for “the Nationals,” support rises to 2.5 per cent (Morgan) or 3 per cent (Newspoll). On these figures, too, support had (at best) almost halved.
How reliable are these figures? Traditionally, the National Party has been underreported by the polls. Not since 2019 have any polls reported a separate National Party figure in their final pre-election forecasts; and in 2019 they all underestimated how well the Nationals would do.
Analysing a post-1966 election survey, he had conducted in 1967, the political scientist Don Aitkin discovered that nearly half the respondents who had voted for a Country Party candidate claimed to have voted Liberal despite there being no Liberal candidate in their electorates. How much of the apparent change since 2025 reflects the same phenomenon — respondents saying they intend to vote Liberal when they will actually vote for a National (the Country Party’s successor) — and how much is due to One Nation eating the Nationals’ lunch, we cannot say.
The National Party might be the party most at risk of losing the greatest proportion of its seats to One Nation, but even if One Nation has already devoured half (or more) of the Nationals’ vote, this would still only account for a small part of One Nation’s support. If One Nation’s vote share has jumped from 6.4 per cent in 2025 to 26.5 per cent (the average of the most recent Morgan, Newspoll, RedBridge and YouGov polls, out of the field by 14 February) and the National’s vote has slumped from 5.6 per cent, for argument’s sake, to 2.5 per cent — a fall of about three points — its losses cannot account for more than about 15 per cent of One Nation’s twenty-point gain.
By far the largest slice of One Nation’s lunch is coming not from the Nationals but from the Liberals. Proportionally, the Liberal losses since the 2025 vote might not be as great as the Nationals’, but in terms of absolute numbers they have been much greater.
In 2025, the Liberal Party secured 20.7 per cent of the vote. During the most recent Coalition split, this dropped by about a third to 14 per cent (YouGov) or 13 per cent (RedBridge) — a fall of seven or eight points. If we add the “Liberal” vote in Queensland, the Liberals’ national vote appears to have dropped by at least a third, from 26.1 per cent in 2025 to 18 per cent (Morgan) or 15 per cent (Newspoll) — losses of between eight and eleven points. In other words, the Liberals alone have contributed 35 to 40 per cent (excluding Queensland) or 40 to 55 per cent (including Queensland) of One Nation’s rise.
Some of those voters who previously supported independents and a variety of very small parties are likely also to have switched to One Nation. In 2025, these voters accounted for 15.2 per cent of the vote. But now, in the most recent polls, independents and others account for no more than 10 to 11.5 per cent of the vote. Those who voted for independents and others in 2025 (though probably not the ones who voted for the teals) could account for 15 to 25 per cent of the rise in the One Nation vote.
Some on the political left have also swung to One Nation. Since we are dealing with net shifts — not the panel data required for us to see all the movement across the parties — the numbers that have shifted from left to right or from right to left can’t be calculated; to the extent that movements in opposite directions have been of equal size, they will have cancelled out. The Australian’s Simon Benson insists that after the Bondi massacre, Labor voters “disgusted” with Albanese or struggling with the cost of living jumped straight over to One Nation. But those losses, he says, were offset by moderate Liberals “appalled by the prospect of the Coalition chasing rabbits down the One Nation burrow.” This may be true, but he provides no evidence for it.
In net terms, the evidence is clear: the boost Labor voters have given One Nation has been relatively small. In 2025, Labor won 34.6 per cent of the vote; averaging the figures in the most recent polls by Morgan, Newspoll, RedBridge, Resolve and YouGov, Labor’s share stands at 32.2 per cent. That drop of just over two percentage points would have contributed about 12 per cent of the surge in One Nation’s support.
If any Greens have moved to One Nation, their numbers would have been much smaller. In 2025, the Greens won 12.2 per cent of the vote; in the latest polls, the Green’s average share (11.7 per cent) remains almost the same.
In short, the lion’s share — as much as 85 per cent— of One Nation’s support appears to have come from the Liberal, National, and Liberal National parties and from smaller parties and independents.
2PP or not 2PP?
More challenging for the pollsters than deciding whether Liberal and National support should be reported separately during the most recent split was deciding what to do with the two-party-preferred, or 2PP, the centrepiece of election reporting since all the polls adopted it in 1993.
To calculate the 2PP, pollsters take all the first preferences for Labor (or for the Coalition parties) and add all the other ballots that rank Labor ahead of the Coalition (or the Coalition ahead of Labor). Labor’s 2PP is then the percentage of all ballots where Labor is ranked ahead of the Coalition (and the Coalition’s 2PP is the percentage of the ballots where it is ranked ahead of Labor).
The concept of the 2PP dates from the early 1970s when: (a) Australia had a system of compulsory and exhaustive preferential voting (as it still does); (b) to win, candidates had to secure more than half the votes, either before or after the distribution of preferences (still the case); and (c) the winner in each seat, almost invariably, was either a Labor, Liberal or Country Party candidate. The third of these factors hasn’t held since around 2013: so much so that in 2025 no fewer than thirty-five seats (known as non-classic seats) out of 150 featured some other party or an independent in the final two-candidate count.
The significance of the 2PP, nationwide, is threefold. First, it shows whether Labor or the Coalition is the more likely to form government. Second, it suggests how many seats Labor would lose to the Coalition, or the Coalition to Labor, given a national 2PP swing of a given size — regardless of how many seats other parties and independents might win. Third, it propagates a new yardstick by which to judge which side deserves to form government — the side with more than half of the 2PP.
In each of these senses the 2PP has succeeded, largely. Its record of picking the winner, though not perfect, has been pretty good: two misses (1998, 2007) in the last 21 elections. Its anticipation of the net shift in seats between the two sides — despite the rise of the Greens and the teals — has been very good; even in 2022 and 2025 it was out by only two seats. Equally successful has been its elevation of the 2PP as the pre-eminent measure of electoral fairness in relation to the distribution of seats.
This last helps explain why calls for proportional representation to replace single-member electorates are largely a thing of a distant past. When Labor, with 54 per cent of the 2PP in 2025, won 63 per cent of the seats with less than 35 per cent of the first preferences, it provoked barely a whisper in favour of switching to PR. It helps, of course, that having succeeded in single-member electorates, the Greens and independents have no great interest in PR; their concerns about electoral reform are focused on electoral funding.
But what to do with the 2PP when the Coalition collapses? On this, pollsters were divided. For some, the collapse was either a temporary separation or, if a divorce, something to be recognised in polling methods sometime in the future. And even before the splits, not all pollsters had been calculating — or had ever calculated — the 2PP figures in the same way.
Morgan continued to provide two figures — a 2PP based on the actual distribution of preferences at the 2025 election, and a 2PP based on how respondents said they would allocate their preferences. (Come the election, this would allow Morgan to highlight whichever measure had come closest to anticipating the result — usually, the distribution at the last election, thanks to voters following their how-to-vote card).
During the parties’ separation, Morgan conducted three polls. Based on the 2025 vote, Labor’s 2PP started at 54.5 before falling to 53 per cent; based on respondent allocation, Labor’s 2PP started at 56.5 before falling to 53.5 per cent. Rather than showing the Coalition losing ground following the separation, as one might expect, these figures showed it gaining ground. A week after the Coalition had reformed, Morgan’s 2PP had it losing ground, with Labor’s 2PP jumping to 58.5 per cent (based on the 2025 allocation) or 55 per cent (respondent allocation) before falling back to 55 per cent (whether based on both the 2025 allocation and respondents allocation was not disclosed).
RedBridge, polling for the Australian Financial Review in association with Accent Research, also continued to calculate a 2PP while the parties were apart, in its case based on preference flows from the 2025 election. It warned, however, that with the new spread of primary votes, “the traditional two-party preferred measure” made the 2PP likely to be “less useful than it has been historically.” Its poll put Labor on a 2PP of 56 per cent. YouGov’s 2PP, also based on the 2025 distribution, was 55–45.
Resolve, polling for the Age and Sydney Morning Herald, would report the same figures, 55–45, the week after the Coalition had got back together. Its 2PP was based on how respondents said they would allocate their preferences. But, as the mastheads’ chief political commentator James Massola was at pains to emphasise, Resolve held back from “highlighting” these figures: “those numbers are potentially inaccurate because in some seats where One Nation is polling well, a two-candidate preferred count that pitched Labor against One Nation, or the Coalition against One Nation, would be more accurate.”
In seats of that kind, this was no doubt true. But Massola appeared to misunderstand what a 2PP measures nationally, and Resolve’s pollster appeared to share a similar misunderstanding. A 2PP, Jim Reed was reported as saying, “only works at a national level when you have a binary contest.” At a national level, despite the large loss of Coalition and Labor votes in recent elections, the 2PP has continued to work well. How well it works with a resurgent One Nation remains to be seen, but no commentator expects Labor to lose the next election or for its 2PP lead (as the Resolve poll indicates), if it wins the election, to be anything other than substantial.
Newspoll, the poll most closely watched by politicians, had previously used “preference flows at previous elections”— not just the flow at the last federal election, apparently — to calculate a 2PP. After the Coalition’s separation, Pyxis thought a 2PP no longer applicable and didn’t calculate one. Regarding its absence as self-explanatory, the Australian offered no justification. Essential, whose polls are reported in the Guardian, also appears to have dropped the 2PP (which it usually bases on respondent allocation) — also, without saying why.
A new 2PP and a 3PP
As well as a standard 2PP, YouGov (in a relatively new partnership with Sky News) produced another quite different two-party vote — one that forced respondents to choose not between Labor and the Coalition but between Labor and One Nation. This had Labor ahead, 57–43. According to the announcement of the poll, the new two-party vote meant that One Nation would become the opposition party… if an election were held today.”
It meant no such thing. In a two-horse race, the horse that doesn’t finish first is indeed likely to finish second. But what YouGov missed is that a party needs to have the second-largest share of seats to form the official opposition, not the second-largest share of an arbitrary two-party vote. Even if One Nation outpolled the combined Liberal and National vote, it wouldn’t necessarily out-seat them.
For Benson, pollsters who had started polling a Labor versus One Nation two-party vote were “premature at best and simply whacky at worst.” Imagining that most seats would come down to this sort of contest would indeed be “whacky,” but wondering which way voters would jump given a choice like this was neither premature nor whacky.
The bigger story, entirely missed, was just how close the YouGov’s new two-party figure was to the traditional Labor–Coalition result — and what that said about how close nearly half the voters polled (especially Liberal and National voters) felt they were to One Nation. An analysis based on the 1998 Australian Election Study showed that what was distinctive about those who gave One Nation their first preference was their alienation from politics, their attitudes to Aborigines and their hostility to immigration. Yet large numbers of Coalition voters, if forced, said they would choose a party that Benson described as “extreme right-wing” over Labor, a party in the mainstream.
Newspoll also wondered about what voters would do if they had to choose between Labor and One Nation. But instead of posing a simple choice of second preference, Newspoll gave respondents three options: “Pauline Hanson’s One Nation,” “Labor,” and “Will just follow how to vote cards/don’t know.” Half (50 per cent) chose Labor, 29 per cent chose One Nation, and 21 per cent said they would either follow the how-to-vote card or didn’t know. For its headline (on the front page), the Australian chose “One Nation a No-Go on Preferences.” This was misleading. In addition to the 29 per cent who said they would preference One Nation, 21 per cent were open to the possibility of doing so — if instructed to do so by the how-to-vote card; that’s 50 per cent. Of the Nationals, 70 per cent said they would preference One Nation with another 24 per cent open to doing that if instructed to do so. Most Liberals, too, said they would either preference One Nation (43 per cent) or might if instructed to do so (24 per cent). Of those who intended to vote for “others”, half (53 per cent) said they would preference One Nation or were open to it. Only among the Greens was a One Nation preference off the table: 91 per cent said they would preference Labor.
Three weeks before the split, when support for One Nation had climbed into the low twenties, Fox & Hedgehog computed a 2PP “match-up” between Labor and One Nation and another between the Coalition and One Nation. Asked to choose between Labor and One Nation, 56 per cent chose Labor and 44 per cent One Nation — figures almost the same as YouGov’s. Fox & Hedgehog had the standard two-party vote at 55–45, identical to YouGov’s.
Freshwater, a few days before the split, produced a standard 2PP as well as a Labor versus One Nation two-party vote. These, too, were very similar: 57–43, Labor versus the Coalition; 53–47, Labor versus One Nation. Of those, in 2025, who had given Labor their first preference, 90 per cent preferred Labor to One Nation; so, did two-thirds of those who had given their first preference to the Greens. All the One Nation voters, three-quarters of LNP voters and most of the Others from 2025 preferred One Nation to Labor. These figures point to a clear left–right divide, voters preferring to switch between parties on the same side, where possible, rather than switch to parties on the other side.
Fox & Hedgehog also calculated a “three-party preferred” taking in Labor, the Coalition and One Nation. Its figures were intriguing: Labor, 46 per cent (building on a first preference of 29 per cent); the Coalition, 29 per cent (25 per cent); One Nation, 25 per cent (21 per cent). So, when asked to choose, most of the Greens (14 per cent of the sample) and Independent or other voters (11 per cent) chose Labor (up seventeen points) with the remainder dividing evenly between the Coalition (up four) and One Nation (up four).
When asked in the same poll to choose between the Coalition and One Nation, respondents chose the Coalition, 63–37. Here, switching between parties on the same side was possible for voters on the right — the two alternatives were drawn from the right — but not for voters on the left. Evidently, Labor voters overwhelmingly chose the Coalition.
How likely is it that Labor voters will be forced to choose between the Coalition and One Nation? Not very. Except for the odd seat in which Labor might come third and the first and second places are filled by one of the Coalition parties and by One Nation, it won’t happen.
If One Nation secured 25 per cent of the vote, it is conceivable that it would pick up something like a quarter of the seats, given that in many seats its vote would be much greater than this. A large One Nation vote that yielded only a small number of seats could well fuel a debate about the fairness of the electoral system — a debate that hitherto has been muted.
What is important about a Coalition–One Nation or a Labor–One Nation two-party vote is what it tell us about the attraction of One Nation to Coalition voters but not to Labor voters. While Coalition voters seem willing to preference One Nation over Labor (unless, perhaps, they are handed how-to-vote cards offering advice to the contrary), Labor voters (and Greens) overwhelmingly prefer the Coalition to One Nation; unless instructed otherwise, Labor voters might assist Coalition candidates threatened by One Nation to hold on.
None of the polls have tested a two-party vote between the Coalition and the Greens. Such a choice would test the willingness of voters, especially Labor voters, to choose the Greens over the Coalition. The result, very likely, would prove to be at least as one-sided as the equally unlikely match-up in the Fox & Hedgehog poll between the Coalition and One Nation.
The party spectrum and voter realignment
While the reuniting of the Liberal and National parties has solved, at least for the time being, some of the problems that have confronted pollsters recently, it hasn’t solved all of them; in particular, it has left them with a series of questions about how to treat the One Nation vote and whether the two-party preferred will continue to work in the way that it has worked for more than fifty years.
With the decline in support for the parties of government since 2019, Australian politics — until the rise of One Nation — had come to be discussed in terms of three more or less equal sized groups: Labor, the Coalition, and an inchoate melange of others. Descriptions of the electoral landscape of this kind have limited analytical value.
Commentators who thought the large group of “others” — those who didn’t intend to vote for either Labor or the Coalition — were just “protest” voters never explained why the same label couldn’t be applied, just as easily, to some of those voting for the parties of government. After the 1993 election — an election in which Labor and the Coalition toegther secured over 89 per cent of the vote — when Newspoll asked respondents which was the “stronger influence” on the way they had voted — “your liking of the party you voted for” or “your disliking of the other parties” — the majority of those who had voted for Labor and the majority of those who had voted for the Coalition chose their “disliking of the other parties.” Known to political scientists as negative partisanship, it’s not something Newspoll has sought to document after recent elections.
Again, commentators who said those in the polls who intended to vote for “others” were simply “parking” their vote — a term Sol Lebovic, who used to run Newspoll, would use regularly — while making up their minds about which way to jump, never stopped to explain why some were “parking” their vote with the Greens while others were “parking” their vote with One Nation. These choices are quite different and not random.
The rise of One Nation has made the idea of three groups of roughly equal size — Labor, the Coalition, and an undifferentiated group of others — even less tenable. The parties that are now attracting about 90 per cent of the vote are the Greens and Labor on the left, and the Coalition and One Nation on the right. The teals, often characterised as fitting somewhere in the middle — to the right of the Labor Party but to the left of the Coalition — hold a much greater proportion of the seats than they hold of the votes.
The surge in support for One Nation, whether temporary or not, and the willingness of Coalition voters to preference it over Labor, challenges what Benson sees as “the great equaliser” of the Australian electoral system — “compulsory preferential voting.” Neither compulsory voting nor exhaustive preferential voting protects the duopoly (Labor and the Coalition) or guarantees that politics stays in the “middle.”
The movement of voters within blocs — left and right — rather than between blocs is not unique to Australia: recent research in Britain points to much the same thing. A world in which most of those who swung between elections moved from Labor to the Coalition or from the Coalition to Labor may have passed. A new world in which voters move between the Coalition and One Nation, or between Labor and the Greens, helps explain why the turmoil of the last few weeks — not least in the polls — has left the 2PP, with Labor well ahead, largely undisturbed. •
