How we read Australian polls
How we turn Australia's polls into a forecast for the House of Representatives (150 seats, 76 for a majority): the 12 polls of the last 30 days, corrected for how Australia's final polls have missed before. Today's average: One Nation 26.6, Labor 26.1, Coalition 23.2, Greens 13.4.
Our rule everywhere: measure how far final polls missed at past elections, take off half the average miss (λ = ½), and keep the correction only if it would have made past forecasts better when each election is predicted without knowing its result. Then run the figures through the country's own voting system, which in Australia means preferences.
Why we don't correct
Averaged over the House elections of 2007–2025 for which final primaries survive (7 elections; the model uses none before 2000, so 1987, 1990, 1998 are left out), the final polls missed the primary vote by: Labor +0.7 (7 elections), Coalition −0.6 (7 elections), Greens +0.8 (7 elections), One Nation +0.9 (3 elections), Others −1.5 (3 elections) points (positive = too high). The λ rule would add Labor −0.36, Coalition +0.29, Greens −0.38, One Nation −0.47, Others +0.77.
Predicting each election with the correction learned from the others was closer at only 2 of 7 elections, and worse on average (1.35 points per party against 1.33 for the plain polls). Australian polls miss, but not the same way twice, so our projection uses them as they are. The test →
The pollsters' own 2PP is not corrected (the forecast works from primaries): their final 2PP missed by 2001 −2.0, 2004 +0.7, 2007 +0.9, 2010 +3.7, 2013 +0.3, 2016 −0.2, 2019 +2.9, 2022 −0.5, 2025 −2.9, an average of +0.3 with no steady direction.
What decides the seats
- Who makes the final two. With One Nation near Labor on primaries and ahead of the Coalition, the count in many divisions ends Labor v One Nation or Coalition v One Nation. Who is excluded third decides the seat.
- Where preferences go. Greens preferences go about nine to one to Labor; Coalition preferences favour One Nation over Labor, but less than One Nation's go to the Coalition.
- The 2PP swing in the divisions that stay Labor v Coalition.
- 76 seats. Anything short is a hung parliament; independents and the Greens then decide who governs.
- 36 final polls with primaries (2007, 2010, 2013, 2016, 2019, 2022, 2025) and 42 with a 2PP (2001, 2004, 2007, 2010, 2013, 2016, 2019, 2022, 2025) at the elections the model uses, from 2000 on. We also hold 3 with primaries and 4 with a 2PP from 1987, 1990, 1993, 1996, 1998, one to three per election; they are shown in the track record but not used by the model.
- AEC results by candidate for every division, 2004–2025, with the two-candidate-preferred and 2PP counts, and the full distribution of preferences: 3,420 transfers in 826 exclusions in 2025 alone, 6,157 since 2004.
- Division results 1984–2001 (first preferences and two-candidate-preferred) from the Parliamentary Library's Parliamentary Handbook; national results back to 1949.
- The AEC's notional 2022 results on the 2025 boundaries (its seat-status fact sheet), the base for the 2025 swings.
- 192 poll averages at set distances before polling day (2010–2025), for the election-day model.
- The 2021 Census on the 2025 boundaries (ABS).
- Average: each pollster's latest national poll within 21 days of the newest, equal weight per pollster; primary votes for Labor, the Coalition, One Nation, the Greens and others (independents included), undecided taken out, rescaled to 100.
- Correction: none: tested on past elections, the polls as published did at least as well without one.
- 2PP: Labor's primary plus each other group's primary times the share of that group's votes that reached Labor in the 2025 2PP count (AEC flow file): Greens 88%, One Nation 26%, others 55%, Coalition under 1% (three-cornered contests). On the 2025 primaries this gives 55.2, the AEC's figure. Today: 50.5.
- Divisions: every 2025 candidate's first-preference share moves by half a uniform and half a proportional swing of their group (One Nation from 6.4% to over 25% would otherwise land either evenly everywhere or only where it was strong). Independents and minor parties move with "others". Where two or more pollsters published a state breakdown in the last 90 days, half the state's swing beyond the national swing is added.
- The count: the lowest candidate is excluded and their votes pass to the others by preference weight (next tab), until two remain.
- Classic seats: if the final two are Labor and the Coalition, the winner comes from the division's 2025 2PP plus the national 2PP swing (plus the state 2PP swing). Otherwise the count decides.
- Chances: 5,000 simulations; each party's national primary moves by its typical final-poll miss (root mean square since 2007), each state by 1.5 points more, each candidate by (1 + 0.08 × share) points, and each classic seat's 2PP by 2.5 points.
- Polling day: the same, with drift for the days left: rmse(days) = 2.17 + 0.19 × ln(days / 30), and each party pulled part of the way back to its 2025 result, as the tests advised.
- Lenses: each past election's average final-poll miss taken off today's polls, and every division run again.
When a candidate is excluded, where do their votes go? We fitted one weight for each pair of groups on all 826 exclusions in the AEC's 2025 distributions of preferences (a Luce choice model: each group still in the count gets a share proportional to its weight). The largest weight from each group is 1.
| From \ to | Labor | Coalition | Greens | One Nation | Independents | Right-leaning minor parties | Others |
|---|---|---|---|---|---|---|---|
| Labor | 0.61 | 0.31 | 0.96 | 0.21 | 1.00 | 0.72 | 0.61 |
| Coalition | 0.30 | 1.21 | 0.11 | 1.00 | 0.68 | 0.08 | 0.12 |
| Greens | 0.93 | 0.21 | 0.56 | 0.19 | 1.00 | 0.25 | 0.52 |
| One Nation | 0.19 | 0.60 | 0.18 | 0.37 | 0.49 | 1.00 | 0.21 |
| Independents | 0.86 | 0.56 | 0.80 | 0.68 | 1.71 | 0.69 | 1.00 |
| Right-leaning minor parties | 0.24 | 0.40 | 0.18 | 1.00 | 0.43 | 0.44 | 0.44 |
| Others | 0.84 | 0.31 | 0.73 | 0.86 | 1.00 | 0.74 | 1.12 |
Groups: independents include Centre Alliance; "right-leaning minor parties" are those whose voters gave the Coalition more preferences than Labor (under 42% to Labor in an AEC 2PP flow file), plus Katter's party, the UAP, Trumpet of Patriots, Family First and similar; others are the rest. Weights between two candidates of the same group (say two independents) are the diagonal. How often Coalition voters put One Nation ahead of Labor rests on the few 2025 seats where both were left after the Coalition was out; the pollsters' own Labor v One Nation 2PP figures suggest a weaker flow than ours.
How stable are the flows?
The share of each group's votes that reached Labor in the AEC's 2PP count, election by election. We use the last election's; the Greens' flow has been steady (and rising), One Nation's has swung towards the Coalition since 2016.
| Election | Greens → Labor | One Nation → Labor | Others → Labor |
|---|---|---|---|
| 2004 | 81% | 44% | 44% |
| 2007 | 80% | 47% | 45% |
| 2010 | 79% | 45% | 42% |
| 2013 | 83% | 45% | 47% |
| 2016 | 82% | 50% | 49% |
| 2019 | 82% | 35% | 46% |
| 2022 | 86% | 36% | 50% |
| 2025 | 88% | 26% | 55% |
The same weights fitted separately on each election's distribution of preferences, read as: when only the two named groups are left, the share going to the first.
| Election | Exclusions | Grn→ALP v Coal. | ON→Coal. v ALP | Coal.→ON v ALP | Ind→ALP v Coal. | Oth→ALP v Coal. | Right→Coal. v ALP |
|---|---|---|---|---|---|---|---|
| 2004 | 791 | 78% | 58% | 42% | 43% | 56% | 71% |
| 2007 | 754 | 76% | 59% | 77% | 50% | 55% | 66% |
| 2010 | 549 | 76% | 59% | 82% | 46% | 49% | 69% |
| 2013 | 888 | 79% | 54% | 70% | 50% | 56% | 56% |
| 2016 | 694 | 79% | 50% | 74% | 53% | 54% | 67% |
| 2019 | 754 | 77% | 66% | 43% | 50% | 74% | 69% |
| 2022 | 901 | 80% | 66% | 64% | 57% | 77% | 66% |
| 2025 | 826 | 82% | 76% | 77% | 61% | 73% | 62% |
One Nation stood in few divisions from 2007 to 2016, so its figures for those years rest on few exclusions. Pooling every election's exclusions instead of using the last one's was tested below: 3 more winners right over 5 election pairs, a small gain, and it would pull today's One Nation flows back towards their 2004–16 levels, so we keep the last election's.
The division model
| Test | Input | Winners right | 2CP miss | Seats |
|---|---|---|---|---|
| The 2025 count replayed from the actual first preferences | 2025 preference weights (in sample) | 149 / 150 | 1.35 (147 pairs right) | Labor 94, Coalition 42, Independents 11, Katter's Australian Party 1, Greens 1, Centre Alliance 1 |
| The 2025 count replayed from the actual first preferences | 2022 preference weights (out of sample) | 146 / 150 | 1.47 (143 pairs right) | Labor 93, Coalition 40, Independents 13, Greens 2, Katter's Australian Party 1, Centre Alliance 1 |
| 2022 → 2025 on the notional 2022 results | the actual 2025 national vote (perfect polls); 2PP swing in classic seats | 133 / 150 | Labor 89, Coalition 46, Independents 9, Others 2, Greens 2, Katter's Australian Party 1, Centre Alliance 1 | |
| 2022 → 2025 on the notional 2022 results | the actual 2025 national vote; the count alone in every seat | 131 / 150 | Labor 83, Coalition 52, Independents 9, Others 2, Greens 2, Katter's Australian Party 1, Centre Alliance 1 | |
| 2022 → 2025 on the notional 2022 results | the actual 2025 national and state votes (perfect state polls); 2PP swing in classic seats | 134 / 150 | Labor 93, Coalition 41, Independents 9, Others 3, Greens 2, Katter's Australian Party 1, Centre Alliance 1 | |
| 2022 → 2025 on the notional 2022 results | the 2025 final-poll average as published; 2PP swing in classic seats | 128 / 150 | Labor 80, Coalition 55, Independents 7, Greens 4, Others 2, Katter's Australian Party 1, Centre Alliance 1 | |
| 2022 → 2025 on the notional 2022 results | the 2025 final-poll average and the 2022-25 state polls; 2PP swing in classic seats (5 states with 2 or more state polls) | 127 / 150 | Labor 77, Coalition 56, Independents 8, Greens 5, Others 2, Katter's Australian Party 1, Centre Alliance 1 | |
| 2022 → 2025 on the notional 2022 results | the 2025 final-poll average, corrected as the history to 2022 advised; 2PP swing in classic seats | 128 / 150 | Labor 80, Coalition 55, Independents 7, Greens 4, Others 2, Katter's Australian Party 1, Centre Alliance 1 | |
| The 2025 result | Labor 94, Coalition 43, Independents 10, Katter's Australian Party 1, Greens 1, Centre Alliance 1 |
"The 2025 count replayed": the actual first preferences in every division, run through our count; 2CP miss is the average gap between our winner's two-candidate-preferred share and the real one, where we found the right final pair. "2022 → 2025": the AEC's notional 2022 figures on the 2025 boundaries, moved by the national (and state) change, counted with the 2022 preference weights only, against the real 2025 winners. The notional figures leave out parties that did not stand in 2025 (mostly the UAP), which we treat as one right-leaning minor party.
Election to election, 2004–2022
Each election replayed from the one before on the actual boundaries: every division with the same AEC id at both (boundaries may have moved in between, which counts against us), its candidates moved by the national change, counted with the earlier election's preference weights, and scored on the grouping that won. The columns change one choice at a time.
| Elections | Our model | Final polls | Count alone | All proportional | ¼ uniform | ¾ uniform | All uniform | Pooled weights |
|---|---|---|---|---|---|---|---|---|
| 2004 → 2007 | 132 / 149 | 131 / 149 | 131 / 149 | 132 / 149 | 132 / 149 | 132 / 149 | 132 / 149 | – |
| 2007 → 2010 | 127 / 147 | 126 / 147 | 128 / 147 | 127 / 147 | 127 / 147 | 127 / 147 | 127 / 147 | 128 / 147 |
| 2010 → 2013 | 132 / 150 | 132 / 150 | 129 / 150 | 132 / 150 | 132 / 150 | 132 / 150 | 131 / 150 | 132 / 150 |
| 2013 → 2016 | 132 / 149 | 132 / 149 | 131 / 149 | 132 / 149 | 132 / 149 | 132 / 149 | 132 / 149 | 132 / 149 |
| 2016 → 2019 | 132 / 143 | 119 / 143 | 132 / 143 | 132 / 143 | 132 / 143 | 132 / 143 | 132 / 143 | 134 / 143 |
| 2019 → 2022 | 129 / 149 | 131 / 149 | 128 / 149 | 129 / 149 | 129 / 149 | 129 / 149 | 129 / 149 | 129 / 149 |
| Total | 784 / 887 | 771 / 887 | 779 / 887 | 784 / 887 | 784 / 887 | 784 / 887 | 783 / 887 | – |
"Our model": half uniform, half proportional swing, the 2PP swing where the final two are Labor and the Coalition. "Final polls": the same with the final-poll average in place of the true change (parties the polls didn't name keep the true change). "Count alone": no 2PP swing, the count decides every seat. The uniform share barely matters in these years, when no party's vote moved far; it matters now, with One Nation's quadrupled. Pooled weights: all earlier elections' exclusions in one fit (not possible from 2004).
The 2PP from primaries
Each election's actual primaries converted with the previous election's flows: 2007: 52.7 against the AEC's 52.7; 2010: 50.4 against the AEC's 50.1; 2013: 45.5 against the AEC's 46.5; 2016: 49.4 against the AEC's 49.6; 2019: 49.3 against the AEC's 48.5; 2022: 51.2 against the AEC's 52.1; 2025: 54.8 against the AEC's 55.2. This is the part of a 2PP miss that comes from flows changing, not from the polls.
Correcting final polls (λ = ½)
Each election is predicted with the correction worked out from the other elections only, so it never sees the result it is tested on. Correcting Australian polls was closer at 2 of 7 and further out at 5 (average miss per party 1.35 against 1.33 for the polls as published). So our projection leaves these polls as they are.
Elections before 2000 (1987, 1990, 1998) are not in this test: they are shown in the track record but not used by the model.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2007 | 0.77 | 1.09 | Correction further out |
| 2010 | 2.74 | 2.59 | Correction closer |
| 2013 | 1.06 | 0.75 | Correction closer |
| 2016 | 0.33 | 0.47 | Correction further out |
| 2019 | 1.38 | 1.45 | Correction further out |
| 2022 | 0.97 | 1.02 | Correction further out |
| 2025 | 2.06 | 2.06 | Correction further out |
Pulling back towards the last result
Closer than the polls as they stand, so used on the election-day page. Details →
No ban on publishing polls
Australian federal law does not stop anyone publishing an opinion poll, at any time, including on polling day. We checked the official sources ourselves on 27 September 2026:
- The Australian Communications and Media Authority, which enforces the broadcasting rules: "You cannot broadcast any election or referendum ads from the end of the Wednesday before the relevant polling day until the close of the poll on polling day", and "Online services (including streamed content viewed on smart TVs) and print media can still publish election or referendum ads during a blackout period" (acma.gov.au, Election and referendum blackout periods). The rule is in the Broadcasting Services Act 1992, Schedule 2, and covers paid election advertising on radio and television, not news reports.
- The AEC's pages on electoral communication deal with authorisation of electoral material and say nothing of a restriction on polls.
So Australia is not in our list of pre-election poll bans, and the Australian figures are never frozen. (The wording above is ACMA's own summary of Schedule 2.)
- Boundaries. We use the 150 divisions of the 2025 election. South Australia, Tasmania and the ACT are being redrawn: the AEC began on 12 August 2025 and its determination is due on 8 October 2026 (Grey is proposed to become O'Donoghue, and Franklin Tongerlongter). Queensland's redistribution began on 28 July 2026, with a determination due on 22 September 2027. Each state keeps its number of seats, but the AEC's notional results for the new boundaries are not yet published, so none of these changes is in our figures.
- Candidates. We keep the 2025 field and move parties, not people. Members who changed party since 2025 are not modelled: Barnaby Joyce (New England) now sits for One Nation, and Zali Steggall and Allegra Spender for Community Strong Australia; their 2025 votes stay with the party they stood for then. New independents are not foreseen.
- One Nation. Its vote has quadrupled since 2025. How that spreads across divisions (we assume half uniform, half proportional) and how Coalition voters rank it against Labor are the two biggest uncertainties, and the 2022 → 2025 test cannot check them, because One Nation hardly moved then.
- State polls. The state swing is kept, as a small adjustment, although in the 2025 test the 2022–25 state polls did not improve the seats; see Tests.
- History. The correction test rests on 7 elections (2007, 2010, 2013, 2016, 2019, 2022, 2025): the model uses none before 2000, and 2001 and 2004 have a 2PP only. Final polls before 2019 do not name One Nation; the 2PP flows come from the AEC's files, which start in 2004.
- Senate. A group-level simplification of the real count; tested on 2025.
Figures as of . Published by Polling Forecast.




