Understanding Australian polls
How to read Australian opinion polls, from house effects and MRPs to who paid for the poll
By Ben Raue, Electoral analyst
Opinion polls are one of the main ways to read Australian politics, and one of the easiest to misread. Ben Raue explains how polls work in Australia: why the trend across many polls matters more than any one result, what reporting should tell you about a poll, who the credible pollsters are, and how federal, state, seat and MRP polls differ.
Polls have had a fair record since the 2019 miss. Most state elections have been called correctly, and in 2025 the polls picked a Labor win while significantly understating Labor’s support. But transparency is still lacking, several prominent pollsters sit outside the Australian Polling Council, and herding can make polls agree even when they are all wrong. Read them with some uncertainty in mind, rather than ignoring them.
The full report · 15 pages, with every chart.
Get the full reportWhat we found
1. The trend across all polls matters more than any single poll
A single poll can show an unusual shift that no other poll repeats, so it is better to look at polls collectively than at one outlet’s latest result. Raue points to trackers that list every poll and publish averages, including William Bowe’s BludgerTrack, Kevin Bonham, Mark the Ballot and Wikipedia’s election pages. Comparing like with like also guards against the “house effect”, where one pollster is consistently more favourable to a party than another.
2. Who paid for a poll shapes how far to trust it
Polls paid for by media outlets or pollsters are published on a regular schedule whatever the result, which makes them the best for analysis. Polls commissioned by parties, unions, lobby groups or charities are usually published only when the result suits the commissioner, so they should be taken with a grain of salt. A big sample doesn’t make up for an unrepresentative one: self-selected online “polls” can be far larger than credible polls and still be easily manipulated.
3. Almost every federal poll is now run online
Of the 42 federal opinion polls conducted between May and July 2026, all but four were conducted entirely online, through panels whose members’ demographics and location are recorded. Resolve’s monthly poll combines online and telephone polling, and Roy Morgan published one SMS poll just after the May budget. Phone polling now suffers from very low pickup rates, which makes a representative sample hard to reach.
4. One Nation’s rise has made two-party-preferred figures harder to rely on
The two-party-preferred figure sorts voters between Labor and the Coalition, which was the final contest in three quarters of seats in 2025. With One Nation polling very highly in 2026, its value has been questioned: some pollsters have stopped producing it, while others now publish Labor versus One Nation figures. With little historical data behind them, those figures rely more on respondents’ own preference allocations.
What’s in the report
- Introduction p. 3
- It’s never about just one poll p. 4
- Who can you trust? p. 5
- Polling different geographies p. 9
- Preference flows and polls p. 11
- How are polls conducted? p. 12
- Can we trust the polls? p. 13
Thanks. The report has opened in a new tab.
