auspol tracker

Aggregated opinion polling for the next Australian federal election, set against the last twenty. Labor leads the Coalition 52.3–47.7 two-party preferred on implied preference flows (±1.3) – updated 11 October 2026 from 181 published polls across 13 polling houses. Next election due by 20 May 2028.

Two-party preferred

Labor 52.3%, Coalition 47.7% on implied preference flows — the pollsters’ own respondent-allocated figures read Labor 51.0%, Coalition 49.0%

Labor leads by 4.6 points (±2.6 on the lead) on implied preference flows – every poll’s primary votes re-allocated by the preference flows the AEC counted at the 2025 election, so every poll that publishes primaries counts, not only those that file a two-party figure. The pollsters’ own respondent-allocated figures aggregate to 51.0–49.0, and the live chart switches between the two. The aggregate is a sample- and recency-weighted, house-effect-adjusted mean over a 21-day window (7-day half-life), carrying a 95% interval of ±1.3 points on each share from 7 polls. Repeat waves from one house inside a window or a calendar month count for the square root of their number, so three weekly waves count as 1.7, not 3.

Against One Nation the implied reading is Labor 52.5%, One Nation 47.5% (±1.9, counting both the polls’ sampling error and doubt about the flow table). No election night has ever counted a Labor v One Nation finish, so that figure runs on the site’s own estimated preference flows rather than the 2025 election’s flows; the pollsters’ few printed head-to-heads sit beside it on the live chart.

Primary vote

Latest polls

Most recent published national polls – the two-party figures read each poll’s primaries at the 2025 election’s preference flows, so the table compares house to house on one fixed allocation; each wave’s own respondent-allocated figure sits in its breakdown on the archive.

About this tracker

auspol tracker pools every published national voting-intention poll since the May 2025 federal election. The two-party and primary-vote aggregates are weighted means: recent and larger polls count for more, and each pollster’s figure is adjusted for its own lean against the consensus of the houses polling around it – pooled with a 90-day half-life, so a house’s lean tracks its current method rather than averaging its whole history – and shrunk toward zero while the evidence is thin. The lean is measured separately for every measure – a firm that leans one way on the classic two-party is not assumed to lean the same way on a primary share or an ALP-v-One Nation head-to-head – and a matchup too few houses ask is left as a plain monthly average rather than adjusted on guesswork. The leaders’ ratings and national direction run through the same monthly weighting and adjustment; preferred prime minister and the undecided share stay as plain averages, the differences there being a matter of question wording rather than lean. Houses that publish no two-party figure feed the primary-vote and leadership series only. Each poll’s weight rests on its published effective sample where the house publishes one – Newspoll, YouGov, Essential, DemosAU, RedBridge/Accent, and Fox & Hedgehog do, in their Australian Polling Council methodology statements – and otherwise on its raw sample, discounted by 1.6 for weighting and capped at 3,000.

The headline carries a 95% interval – the greater of the spread among polls in the window and their sampling error – currently about ±1.3 points on 7 polls across 21 days (effective sample 4.7 after weighting). It cannot cover error the whole industry shares: an aggregate has no way to see a lean every poll in it carries. Movement smaller than the interval is marked as such.

That caveat is not idle. Across the 18 elections from 1974 to 2025 the final polls missed the two-party result by 1.4 points on average – at 1984 by 3.2, every house on the same side of it. Past cycles carries the full record, house by house.

Reading the charts

On the two-party chart each dot by default is one poll’s implied figure – its primaries re-allocated at the 2025 election’s counted flows – and the switch under the heading swaps the whole series to the pollsters’ own published figures, dots and all. Elsewhere each dot is one published poll. The lines are monthly aggregates, shaded with the 95% interval around them. Where the two bands meet, that month’s lead is inside its own margin of error. Leadership questions are asked irregularly, so those lines are monthly aggregates too – adjusted per house for approval and favourability, joined straight from published readings for preferred prime minister. A “—” in any table means the pollster didn’t ask that question.

Why there is no seat projection here. Turning a national two-party figure into a seat count assumes a uniform swing, and with One Nation near 26% of the primary vote the assumption fails in exactly the seats that would decide the election: a large minor party wins seats where its vote is concentrated and none where it is not – and no national number knows the difference. Seat figures appear on this page only where a pollster modelled them seat by seat and published the result, which is what the MRP tag in the archive marks.

Sources

Roy Morgan, YouGov, Newspoll, RedBridge/Accent, Resolve, Essential, DemosAU, Newspoll (pooled), Freshwater, Spectre Strategy, DemosAU (MRP), Fox & Hedgehog, YouGov (MRP), Wolf & Smith, EMRS (Tas), RedBridge/Accent (MRP), Roy Morgan (SMS), Agenda C Synesis, Roy Morgan (pooled). Field dates and sample sizes are listed per poll in the archive.

auspol tracker is an unofficial aggregator of published federal opinion polling. Best efforts are made to make the aggregate figures transparent, trustworthy, statistically sound, and informative, but they are, in the end, estimates only. Federal polling archives I’ve located are stored here for safekeeping and convenience.