Source: The Conversation (Au and NZ) – By Adrian Beaumont, Election Analyst (Psephologist) at The Conversation; and Honorary Associate, School of Mathematics and Statistics, The University of Melbourne
With almost all primary votes now counted to two-party preferred (as I explained on May 29), Labor has won the national two-party vote by a 55.3–44.7 margin, although this may drop to a 55.2–44.8 margin once the remaining votes from Bradfield come in.
Labor’s two-party share is over two points higher than in any poll taken in the final week before the election.
Final primary votes were 34.6% Labor (up 2.0% since the 2022 election), 31.8% Coalition (down 3.9%), 12.2% Greens (steady), 6.4% One Nation (up 1.4%), 1.9% Trumpet of Patriots (down 2.2% from United Australia Party in 2022), 7.4% independents (up 2.1%) and 5.7% others (up 0.6%).
The table below shows the primary vote and two-party estimates of all ten polls conducted in the final week before the election, with the election results at the bottom. When polls gave a breakdown for Trumpet of Patriots, independents and others, I’ve combined these for an all Others total. Bold numbers in the table represent estimates that were within 1% of the result.
Fieldwork dates for the Ipsos poll were not released, but it was published in The Daily Mail on election day, so it was presumably taken in the last week. Published primary votes in this poll included 5% undecided, which I have redistributed proportionally to the parties listed.
In 2019, all the polls gave Labor between a 51–49 and a 52–48 lead. The actual result was a Coalition win by 51.5–48.5.
This year, all polls had Labor between a 51–49 and a 53–47 lead and the actual result was a Labor win by 55.3–44.7. The two polls (Freshwater and Ipsos) that had Labor below a 52–48 lead were particularly poor.
The polls understated Labor’s primary vote and overstated the Coalition’s. Labor won the primary vote by 2.7 points, when nearly all polls had the Coalition ahead (Redbridge was tied). The Freshwater and Ipsos polls performed badly in overstating the Coalition’s vote.
The Greens were mostly overstated, while One Nation was overstated by every pollster except Morgan.
Preference flow assumptions compounded the polls’ problems. If I plug the election primary votes into my 2022 preference flows spreadsheet, I get a Labor two-party lead of 55.3–44.7, the same as the actual result.
Newspoll had higher One Nation preference flows to the Coalition than in 2022. If they’d used 2022 flows, Labor would have led by about 53–47. YouGov used data from its MRP polls that gave the Coalition both a higher share of One Nation and Greens preferences than in 2022. If they’d used 2022 flows, Labor would have led by 54.2–45.8.
We won’t have data on preference flows by party for some time, but it’s likely that One Nation preferences did become more pro-Coalition. However, Greens and independent preferences compensated by becoming more pro-Labor.
Respondent-allocated polls from Essential, Resolve, Freshwater, Redbridge and Spectre all suggested this would be the case. YouGov may have used MRP polls earlier in the year to allocate preferences. Labor was doing badly on preferences earlier.
The poll graph that I used in my pre-election articles is below. There was a surge to Labor in March and April. Labor had been polling poorly from December to February and may have lost an election held then. The polls told us that Labor had recovered to an election-winning position, but they understated the magnitude of that win.
The best two polls were not the final polls, but a Morgan poll taken two weeks from the election that gave Labor a 55.5–44.5 lead. Morgan’s final two polls both gave Labor a 53–47 lead. The other good poll was a Redbridge poll of 20 marginal seats that gave Labor a 54.5–45.5 lead a week before the election (actual result 54.8–45.2 to Labor across these seats).
Redbridge would have been better if they’d stuck with their 54.5–45.5 to Labor in the marginal seats in this poll, but they dropped back to 53–47 to Labor in the poll published on election day.
The final YouGov MRP poll predicted Labor would win 84 of the 150 seats, understating Labor by ten seats. An exit poll of early voters from the first two days of early in-person voting correctly had swings to Labor.
While public polling was poor at this election, Liberal internal polling was worse. This article in The Australian published the day before the election said the Coalition was confident of gaining ten seats from Labor. Labor actually gained 14 seats from the Coalition.
The worst seat polls
I’m not going to relate every seat poll in this election, but there were some seat poll stinkers.
I referred to JWS seat polls of Ryan, Brisbane and Griffith on April 18. These polls gave the Liberal National Party a 57–43 lead over Labor in Ryan, with the Greens a distant third on primary votes. In Brisbane, Labor led the LNP by 51–49. In Griffith, Labor led the LNP by 51–49, but the LNP led the Greens by 53–47.
In Ryan, the Greens made the final two and defeated the LNP by 53.3–46.7. If Labor had made the final two, they would have won by 57.8–42.2. In Brisbane, Labor crushed the LNP by 59.0–41.0. In Griffith, Labor and the Greens made the final two, and a two-party count between Labor and the LNP had Labor winning by 65.9–34.1.
I referred to a Compass seat poll of McMahon on April 11. This poll gave right-wing independent Matt Camenzuli 41% of the primary vote, the Liberals 20% and Labor incumbent Chris Bowen just 19%. Bowen actually won 45.5% of the primary vote, the Liberals 26.8% and Camenzuli just 9.8%.
I referred to KJC polls of four seats on April 27. These polls gave the Liberals a 49–45 lead including undecided in Tangney and a 46–41 lead in Blair. In Richmond, the Greens led Labor by 39–34. In Hunter, Labor led the Nationals by 45–41.
Labor actually won Tangney by 57.0–43.0 and Blair by 55.7–44.3. In Richmond, the Greens did not make the final two, and Labor would have beaten them easily if they had. In Hunter, One Nation instead of the Nationals made the final two, with Labor winning by 59.0–41.0. Had the Nationals made the final two, Labor would have won by a similar 59.5–40.5.
Recount results and Greens senator defects to Labor
In Liberal-held Bradfield, Teal Nicolette Boele defeated the Liberals by 26 votes after a recount, overturning an eight-vote Liberal lead on the original count. The Liberals could challenge this result in the courts, but Boele will be seated until the courts decide.
In Goldstein, the partial recount of primary votes for Teal incumbent Zoe Daniel and Liberal Tim Wilson was completed on May 31. Wilson won by 175 votes, down from 260 before the recount started.
With these results, the final seat outcome of the election is 94 Labor out of 150, 43 Coalition and 13 for all Others. That’s a Labor majority of 38 by the UK method.
Western Australian Greens Senator Dorinda Cox, who was elected in 2022, defected to Labor on Monday. This gives Labor 29 of the 76 senators and the Greens ten. Labor will still need either the Coalition or the Greens to reach the 39 votes required for a Senate majority. Cox’s six-year term will expire in June 2028.
South Korea and Poland elections
On Tuesday the centre-left candidate won the South Korean presidential election that had been called early after the previous right-wing president was impeached and removed from office. On Sunday the Law and Justice (PiS) candidate won the Polish presidential election, defeating a pro-Western centrist.
Donald Trump’s US national ratings have improved since his nadir in late April. I wrote about these events for The Poll Bludger on Wednesday.
Adrian Beaumont does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
– ref. Final counting shows polls understated Labor in 2025 election almost as much as they overstated it in 2019 – https://theconversation.com/final-counting-shows-polls-understated-labor-in-2025-election-almost-as-much-as-they-overstated-it-in-2019-256981
