Research
How the forecast works, in full
The 11 papers behind our Australia forecast (19,448 words): how far the polls have missed, how we correct them, the seat model and the tests.
These papers set out every part of our method: what we measure, how we correct the polls, how votes become seats, how uncertain the result is, and what we have tested and rejected. Figures marked Live are computed from the database each time the page is built, so they always match the forecast.
Also: France · Germany · Spain · Italy · Netherlands · Poland · Canada · United States · Great Britain
11 papers · about 19k words · Short summary of the method · How we adjust today's polls · Data sources
Australia
- Forecasting Australia's House from imperfect polls: an overview Australian final polls have missed each main party's primary vote by about 1.4 points at the seven House elections since 2007, overstating Labor at five, but a correction learned from those misses would have made our forecast worse, so the projection takes the polls as published. What decides the seats is the preference count: with One Nation level with Labor, most divisions are three-way contests. This paper sets out what the Australia section forecasts, how each step works, where the uncertainty lies, and how the other Australia papers fit together.
- How wrong have Australian polls been? Primaries, the 2PP and the λ test At the seven House elections from 2007 to 2025, the average of the final polls missed each main party's primary vote by 1.4 points. Labor was overstated at five, but in 2025 every final poll had Labor too low. The pollsters' own 2PP missed by 1.6 points on average with no steady direction. We define the error, describe the 46 final polls and their sources, report the gap between the two largest parties, each pollster's record and state elections as a check, and show why the leave-one-out test rejects a correction.
- From primaries to seats: preference flows and the 2PP The forecast works out Labor's 2PP from the primaries with the share of each group's votes that reached Labor in the AEC's 2025 count, which came within about a point of the real 2PP at every election from 2007 when tested with the previous election's flows. Each division's count is replayed with preference weights between seven groups, fitted by a Luce choice model on all 826 exclusions of 2025; on the actual 2025 votes it picks 149 of 150 winners. The weakest number is how Coalition voters rank One Nation against Labor.
- The division model and its tests Every 2025 candidate's first preferences move by half a uniform and half a proportional swing of their group, with half of any state swing shown by two or more pollsters, and the count is replayed; where Labor and the Coalition are the last two, the 2PP swing decides. We set out the formula, its effect on a surging One Nation, a worked example in Dickson, and the tests: 133 of 150 winners from the notional 2022 results with the true 2025 vote, and 784 of 887 replaying each election from the one before since 2004.
- Simulating Australian polling error: where the chances come from The seat ranges, the chance of a hung parliament and the chance that any combination reaches 76 come from 2,000 simulated elections. Each moves every group's national primary by a random error the size of its past final-poll miss, adds state, candidate and classic-seat 2PP errors, and replays every division's count. On the polls of 15 October 2026 no party reaches 76 in 95% of runs. We set out the design, today's outputs, why the runs lean away from One Nation, and what the simulation leaves out, chiefly uncertainty in preference flows.
- Past-miss lenses: today's polls with each past error replayed Each lens takes today's poll average, removes one past House election's final-poll miss, and runs every division again. On the polls of 15 October 2026 none of the seven lenses, 2007 to 2025, gives any party a majority. A repeat of 2025's miss, which understated Labor, gives Labor 75 seats; a repeat of 2022's gives One Nation 71; the 2010 and 2019 misses double the Coalition's seats. 56 divisions change hands under at least one lens. We define the lenses, work through 2019, and compare them with the simulation.
- Polling day: how far Australian polls move We compared polls at nine distances before each House election from 2010 to 2025 with the result and with each party's previous result. Pulling each party part of the way back towards its last result, by a curve fitted on the other elections, came closer than the polls as they stood at every distance, so the election-day forecast does it: 583 days out, today's polls count for 31% of the gap from 2025. That gives Labor 91 seats on polling day. For One Nation, polling four times its 2025 vote, the pull-back is an assumption, not evidence.
- The half-Senate: a simplified quota count The next half-Senate election fills 40 of 76 seats: six per state and two per territory, beside 36 senators who continue to 2031. Each state's 2025 Senate vote moves with the change in the House polls, and seats come from a group-level quota count with 30% exhaustion, which gives 37 of the 40 seats of 2025 to the right party. On today's polls Labor and the Greens would hold exactly 39, the number needed to pass a bill, and reach it in 26% of simulations. We set out the count, the test and its limits.
- Rating Australian pollsters, and how to read a poll release We score each Australian pollster out of 100: accuracy of its final House polls since 2000, ranked against all 36 Australian final polls and shrunk towards a typical poll; its error on the gap between Labor and the Coalition; and what it publishes. Newspoll has the only seven-election record and scores 62, a B. The rating does not weight the average. We also show each pollster's lean at past elections and today, and explain the parts of an Australian release that need care: the undecided share, the 2PP method and state breakdowns.
- The Australian data, its sources and the law on polls Results, preferences and boundaries are official: the AEC for every division from 2004 to 2025, the 2025 Senate and the Farrer by-election; the Parliamentary Handbook for 1984 to 2001 and national results back to 1949; the ABS for boundaries and the Census. Polls are graded by source: of 172 national polls since 2025, 52 are the pollster's own release and 116 come from Wikipedia's tables. Australian law does not restrict publishing polls, so the figures are never frozen. The 2026 redistributions are not yet in the data.
- What we don't know yet: limitations of the Australian forecast The Australian forecast rests on seven elections of final polls, two of them with one or two polls each. It has no precedent for a party at One Nation's level: how its vote spreads across divisions, how Coalition voters rank it against Labor, and whether it keeps its gain to polling day cannot be tested on past elections. The simulation holds preference flows fixed; the division model keeps the 2025 candidates and boundaries; state polls are few; the Senate count is simplified; most polls are second-hand. For each we say what evidence would change our mind.
