Federal MRP, October–November 2025

01

Labor's primary vote falls and its seat count rises

Both major parties lose primary vote against the May 2025 election and One Nation takes most of what they lose. Labor's two-party position improves by 0.8 points and its central seat forecast rises from 94 to 98.

The bars show first preferences at the May 2025 election and in the October–November 2025 MRP model, in per cent.

The central forecasts total 150 and so do the 2025 seats. The report describes the ranges as assessing in-doubt seats as variable within the range. The Coalition's 29 splits into 21 Liberals and 8 Nationals. This model publishes a two-party preferred figure of 56–44 to Labor against the election's 55.2–44.8, a swing of 0.8 points; it is the only one of the three MRPs in this series to publish a national two-party number.

The arithmetic that produces four extra Labor seats from a smaller Labor vote is preference leakage. Labor's primary falls 1.6 points and the Coalition's falls 7.8, so the gap between them widens from 2.8 points to nine even as both go backwards. One Nation's 10.6-point rise then costs the Coalition a second time at the distribution of preferences, because a vote that leaves the Coalition for One Nation does not return to it in full. The report attributes the two-party swing in Queensland in particular to this effect.

The state two-party figures put Labor ahead everywhere: Tasmania 64–36, the Australian Capital Territory 71–29, South Australia 60–40, Victoria 58–42, Western Australia 56–44, New South Wales 55–45, the Northern Territory 54–46 and Queensland 52–48. The report records two-party swings to Labor of 2.6 points in Queensland, 1.6 in Victoria, 0.8 in South Australia and 0.7 in Tasmania.

A caution belongs with the seat table before anything is built on it. These are modelled estimates rather than measured shares, the report publishes no margin of error for any national figure, and the next election was more than two years away when the fieldwork closed. The report says so itself, describing the model as a snapshot of contemporary voting intentions whose most valuable insight is the shape of the One Nation challenge rather than a forecast of an election.

02

Seventeen per cent becomes twelve seats for four reasons, and none of them is the vote

One Nation has contested federal elections since the late 1990s without winning a seat in the House of Representatives. The report sets out what has changed, and the vote share is only the first item on its list.

1934 Twelve seats would be the most significant third-party result in the House of Representatives since Lang Labor took its bloc in that year, and One Nation's best seat share since the eleven it won in the 1998 Queensland election from 23 per cent of the vote.

The first reason is the one everybody names: the vote. Seventeen per cent is more than two and a half times the 6.4 One Nation took in May 2025. The other three are structural, and they are what turn that share into seats rather than into a large national also-ran.

The second is that the Coalition now directs preferences to One Nation on its how-to-vote cards, which it did not do in the late 1990s. Where the Coalition is eliminated before the final count, its flow to One Nation over Labor exceeds 80 per cent. The third is the mirror image: Labor does place the Coalition above One Nation, but where Labor is eliminated it is usually in rural seats it has little chance of winning and where it fields few booth workers, so its flow to the Coalition over One Nation has fallen below 60 per cent. A how-to-vote card only works where someone is handing it out.

The fourth reason is the Coalition's own weakness. At 24 per cent nationally it is low enough that One Nation can finish ahead of it on primaries and then ride Coalition preferences to victory, which inverts the usual relationship between a major party and a minor one. The report adds a fifth consideration: a 13 per cent Others vote remains outside the Greens and One Nation, much of it anti-major-party, and where the final three are One Nation and the two majors those preferences can lift One Nation from third into the final count.

Taken together these say that One Nation's seat total in this model is as much a product of the preference architecture as of its support. That cuts both ways for a reader. It means twelve seats from 17 per cent is not the anomaly it looks like. It also means the figure is sensitive to decisions the parties have not yet made about the next election's how-to-vote cards, and the model assumes the current ones.

03

One 2025 Coalition voter in three is leaving, and one in twenty is not

Across the model's grid of location, age and education, the share of 2025 Coalition voters now intending to vote One Nation runs from 5 per cent to 35. The national figure is about one in five.

These are modelled estimates from the post-stratification frame rather than observed cell counts, and the report publishes no interval for them. The equivalent grid for 2025 Labor voters runs from 1 per cent, among inner-metropolitan postgraduates aged 18 to 34, to 12 per cent among rural and regional voters over 55 who did not finish Year 12.

Every one of the three dimensions moves the figure in the same direction, and they compound. The lowest cell in the table is an inner-metropolitan postgraduate under 35 at 5 per cent; the highest is a rural voter over 55 without Year 12 at 35. That is a seven-fold difference inside a single party's 2025 vote, and it is why the One Nation challenge in this model is geographically concentrated rather than national in the way the headline 17 per cent suggests.

The Labor grid matters as much for what it rules out. Its highest cell is 12 per cent and its national figure is about one in twenty, so Labor is losing roughly a quarter as much of its 2025 vote to One Nation as the Coalition is. The same three dimensions govern it, which says the movement is an attribute of voters rather than of the party they came from; the Coalition is simply holding far more of the voters who have it.

04

Seven of the Nationals' fifteen seats have One Nation in front

The model puts the Nationals ahead in eight of the fifteen seats they won in 2025 and One Nation ahead in seven. Only two of the fifteen are classified firm holds.

The Nationals are the party this model treats worst, and the reason follows directly from section 03: their seats are rural, their electorates have the lowest share of degree holders, and their voters skew oldest. Lyne and New England are the only two of the fifteen the model calls firm. One Nation is a firm winner in two, Capricornia and Flynn, and the remaining eleven are spread across lean and in-doubt classifications in both directions.

The Liberals are exposed in seven seats, with One Nation firmly ahead in Wright, slightly ahead in close contests in Canning, Grey and Groom, and close behind in Farrer, Forrest and O'Connor. In Longman One Nation pushes the Liberals into third and is itself narrowly behind Labor. The rise also costs the Coalition in its contests against independents through preference leakage, which is how Wannon arrives at a rounded 50–50.

These are the seats the report names in its own summary. Per-electorate primary votes for all 150 seats are set out on pages 16 to 23 of the report and the same seats sorted by Labor two-candidate preferred on pages 24 to 31.

Labor is not untouched. Hunter and Forde are in doubt on narrow Labor leads, and One Nation trails closely in Paterson, Leichhardt, Blair, Whitlam and Hawke. The report's summary figure is that One Nation makes the final count against Labor in 26 seats, which is the number that matters for a party planning a campaign rather than reading a headline: the fight is not confined to the Coalition's side of the chamber.

One Nation is also competitive against sitting independents and minor-party members, which the report names as Calare, Indi and Kennedy. And it reaches beyond Queensland: the report records One Nation ahead or competitive in at least one seat in every mainland state. Against all of that, the model makes One Nation the firm leader in only three seats out of 150, with a range of 8 to 18. The twelve-seat central forecast is the middle of a wide and soft distribution, not a settled number.

05

What this model does not know, in its own words

An MRP reports no effective sample and no margin of error on its national figures. This report records both as not applicable and publishes a single uncertainty figure, which applies to electorate vote shares rather than to anything on the national tables.

The stated uncertainty is a 95 per cent confidence interval of about ±5 points on each electorate's vote share. Seats are then classified against that: in doubt where the winning two-candidate preferred is 50 to 51 per cent or a trailing candidate could win on a 0 to 2-point gain, lean at 52 to 54 per cent or a 2 to 4-point gain, and firm for everything else. Nothing in the report supports attaching an interval to the national primary figures in section 01, and this briefing does not compute one.

Three further limitations are the report's own. It models no change of incumbency in any seat, so every estimate assumes the sitting member recontests, which the report notes is a strong assumption more than two years out. The determination of which two candidates reach the final count rests on 2025 preference distribution data that is not static between elections. And where the model produces a two-candidate pairing the 2025 election did not, it borrows flows from the nearest comparable contest.

That last point deserves naming precisely, because it carries the section 02 argument. For Coalition-against-One Nation final counts the model uses the preference flows observed in Maranoa, and for Labor-against-One Nation counts those observed in Hunter. Two electorates are standing in for every contest of their kind in the country, and the report says so, calling it an extra factor of uncertainty. The next model in this series replaces those two proxies with measured data, and ACE-IU-2026-042 sets out what changed.