Australia Next federal electionby 20 May 2028in 546 days Hung parliament 96% Labor majority 1% Calendar Track record Research
Research · Australia

Polling day: how far Australian polls move

Abstract. 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.

Polling Forecast · updated 15 October 2026 · 6 min read · Australia, election day, poll drift, horizon, pull-back, previous result, uncertainty, One Nation, election date

Summary

The projection answers a hypothetical question: who would win if the election were held today. The election-day forecast answers the real one, for a polling day that may be 583 days away. We took the polls at nine fixed distances before each House election from 2010 to 2025, from 30 days to two and a half years, and compared them with the result and with each party's previous result. Months out, the previous result has often been as good a guide as the polls: at 548 days and beyond, the polls missed each party by 2.6 to 3.1 points on average and the previous result by 2.4. Pulling each party part of the way back towards its previous result, by a weight fitted on the other elections, came closer than the polls as they stood at every distance tested. So the election-day forecast does it: 583 days out, today's polls count for 31% of the gap from the 2025 result. On today's polls that gives Labor 91 seats and an 84% chance of a majority. For One Nation, polling four times its 2025 vote, the pull-back treats a surge without precedent like any other mid-term lead. That is an assumption, not evidence, and this paper says so.

The question

Australian polls are published for the whole three-year term. A poll 19 months before polling day measures opinion now; the result measures opinion on the day, after a campaign, events and the ordinary movement of voters back towards the parties they usually support. The election-day forecast needs two things from the record: where the shares are likely to be on the day, given where they are now, and how uncertain that is.

The date

The House first sat on 22 July 2025 and expires three years later. To hold the House election with the usual half-Senate election, polling day must fall between 7 August 2027 and 20 May 2028; a House-only election could be as late as September 2028. The site uses 20 May 2028, the latest date for the usual joint election, which on 15 October 2026 is 583 days away. The Prime Minister chooses the date, and an earlier one would narrow the ranges at once.

The evidence

For each House election from 2010 to 2025, we averaged the polls at fixed distances before polling day: 30, 61, 91, 182, 273, 365, 548, 730 and 913 days (and 1,095 days for 2013 alone). At each point each pollster's polls with fieldwork within 30 days either side were averaged, then the pollsters' averages were averaged. Parties measured are Labor, the Coalition, the Greens, and One Nation from 2019. That gives 192 poll averages, each set against the party's result at that election and its result at the election before. The model uses no election before 2000, and the series starts in 2010; the 2010 series has only Newspoll and Roy Morgan.

The average miss per party at each distance and election:

Days out 2010 2013 2016 2019 2022 2025
30 2.45 1.73 0.79 1.94 1.48 2.10
61 2.54 1.70 0.74 2.67 2.21 2.67
91 1.95 0.91 0.93 2.51 2.60 3.01
182 1.71 1.65 2.71 3.26 1.78 2.57
273 3.71 1.57 1.62 3.31 2.21 2.55
365 4.66 0.99 2.40 2.78 2.61 2.32
548 6.24 1.25 3.11 4.14 2.74 1.27
730 5.38 3.16 2.64 2.76 2.53 1.01
913 7.02 1.77 1.74 2.46 1.92 0.95

Table: The polls' average miss per party, primary vote, points, at each distance before each House election.

The misses grow with distance, but not evenly. 2010 stands out: two and a half years out the polls had Labor on 48.0% and the Coalition on 34.0%; Labor won 38.0% and the Coalition 43.6%. 2025 runs the other way: the polls two years out were closer to the result than the polls three months out, because Labor polled about 37% two years out, about 31% from a year out to the last month, and won 34.6%.

The method

The pull-back

The same curve as for Britain (From today to polling day: how far British polls move before a general election) gives today's polls a weight that falls with distance from polling day, the rest going to the party's previous result:

\[ w(d) = f + (1 - f)\, e^{-d/\tau}, \tag{1} \]
\[ \hat{v}_p(d) = v^{\text{prev}}_p + w(d)\,\big(y_p - v^{\text{prev}}_p\big), \tag{2} \]

where \(d\) is the number of days to polling day, \(f\) is the floor (the weight that remains however far out), \(\tau\) the time scale, \(y_p\) today's average and \(v^{\text{prev}}_p\) the previous result. A party with no previous result is taken as polled. The shares are rescaled to the same total.

The curve is fitted by least squares over every party, distance and election, searching \(f\) from 0 to 1 in steps of 0.01 and \(\tau\) from 5 to 600 days in steps of 5. If the best fit were \(f = 1\), polls would be taken as they stand at every distance.

The test

The pull-back is used only if it beats the polls as they stand out of sample. At each distance, each election is predicted with the curve fitted on the other elections, and the average miss compared:

Days out Elections Polls' miss Previous result's miss Pull-back's miss RMSE used for drift
30 6 1.75 2.40 1.68 2.09
61 6 2.09 2.40 2.02 2.72
91 6 1.98 2.40 1.86 2.71
182 6 2.28 2.40 2.15 2.83
273 6 2.49 2.40 2.32 3.05
365 6 2.63 2.40 2.19 3.15
548 6 3.12 2.40 2.32 3.97
730 6 2.91 2.40 2.01 3.46
913 6 2.64 2.40 2.25 3.69
Mean 2.43 2.40 2.09

Table: Each election predicted with the curve fitted on the other five, average miss per party, points.

The pull-back was closer than the polls at every distance, by 0.07 points at 30 days and 0.8 to 0.9 points at 548 and 730 days. It was also closer than the previous result alone at every distance. From 273 days out, the polls on their own were further out than simply assuming every party would repeat its last result. The fitted curve on all six elections has

\[ f = 0.31, \qquad \tau = 75 \text{ days}. \]

With \(\tau\) = 75 days, \(e^{-d/\tau}\) is negligible beyond about a year, so for most of the term the weight on today's polls is the floor: 31% of the gap between the polls and the last result is kept, 69% is pulled back.

The drift

The uncertainty grows with distance. The root-mean-square miss at each distance (the last column, from the pull-back's own predictions) is fitted as

\[ \text{rmse}(d) = r_{30} + k\,\ln\!\left(\frac{\max(d, 30)}{30}\right), \qquad r_{30} = 2.17, \quad k = 0.19, \tag{3} \]

and the drift used in the simulation is \(\delta(d) = \max(\text{rmse}(d),\ 0.5)\), scaled down linearly inside the last 30 days to zero on the day. At 583 days, \(\delta = 2.17 + 0.19 \ln(583/30) = 2.73\) points. In the polling-day simulation each group's national error has standard deviation \(\sqrt{\sigma_g^2 + \left(\delta\,\sqrt{\max(S_g,1)/20}\right)^2}\), so the drift is 2.7 points for a group at 20%, more for larger groups and less for smaller ones (Simulating Australian polling error: where the chances come from).

The polling-day forecast

At 583 days, \(w = 0.31 + 0.69\, e^{-583/75} \approx 0.310\). Equation (2) on today's average:

Party 2025 result Today Polling day, equation (2)
Labor 34.56 27.67 32.4
Coalition 31.82 21.42 28.6
Others 14.98 11.25 13.8
One Nation 6.40 26.50 12.6
Greens 12.20 13.17 12.5

Table: Primaries on polling day, 20 May 2028, %.

For One Nation, \(6.40 + 0.310 \times (26.50 - 6.40) = 12.6\). Run through the division model and 1,000 simulations with the drift:

Party Seats today Seats on polling day 80% range Majority
Labor 57 91 72–100 84%
Coalition 13 38 23–54 1%
Independents 11 14 8–21 0%
One Nation 64 4 1–14 0%
Greens 4 1 0–5 0%
Centre Alliance 1 1 1–1 0%

Table: The polling-day forecast, 15 October 2026.

A hung parliament falls from 95% today to 15% on polling day, and Labor's 2PP has an 80% range of 50.5 to 58.3. The site shows this beside today's forecast, not instead of it.

The caution about One Nation

The pull-back was fitted on mid-term swings of ordinary size: governments falling in the polls and recovering, oppositions rising and falling back, the Greens moving a few points. The record contains nothing like One Nation's rise from 6.4% to over 25% within a year. Its highest federal vote before 2025 was 8.4%, in 1998. Pulling it 69% of the way back to 2025 is what the history of other parties' mid-term movements advises, and nothing in the record tests it on a party in One Nation's position.

There are reasons to think One Nation's case could differ in either direction. Populist-right parties abroad have often kept much of a mid-term poll gain at the following election; the site's own analysis of 29 mid-term poll rises across nine countries found that populist-right parties kept 82% of their gain at elections, against 26% for other challengers. Applied to One Nation, that would mean a floor far above 0.31. On the other hand, One Nation's last mid-term rise faded: before the 2019 election it polled 8.9% at 548 days out and won 3.1%. The election-day page carries this caution next to the figures. The forecast for today, with no pull-back, is always shown beside it.

What the election-day model leaves out

Party-specific pull-back. Every party gets the same curve. A party whose gain is lasting and one whose gain is temporary are treated alike.

Six elections. The curve and the drift rest on six elections, the longest distances on fewer polls. 2010 alone has misses above 5 points at 548 days and beyond.

The date. The model assumes the latest usual date. An earlier election would raise the weight on today's polls only if it fell within a few months: with \(\tau\) = 75 days, the weight is still 0.31 at a year out and rises to 0.37 at 180 days and 0.62 at 60 days.

Events. The record includes changes of leader, the pandemic and economic shocks, and the curve averages over them. It cannot foresee the next.

References

Australian Electoral Commission, national results, 2007–2025 (CC BY 4.0).

Commonwealth of Australia Constitution Act, sections 13 and 28.

Jennings, W. and Wlezien, C. (2016). The timeline of elections: a comparative perspective. American Journal of Political Science, 60(1), 219–233.

Figures are as of 15 October 2026, 583 days before the latest usual date. The site's pages update with each new poll.

How to cite

Polling Forecast (2026). Polling day: how far Australian polls move. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/australia-election-day. Live figures change as new polls arrive; the date above is when the text was last revised.

You may also like