New South Wales, October 2025

01

The lower-house question offered four parties, and the upper house shows what the fourth was holding

Respondents choosing a lower-house vote could name Labor, the Coalition, the Greens or any other candidate. The upper-house question, asked immediately afterwards, listed sixteen parties and candidates, and One Nation took 15 per cent of it.

The lower-house figures exclude 6 per cent of voters who were undecided, the upper-house figures 12 per cent. The upper-house column sums to 100. Lower-house undecided voters were asked a leaner question first; upper-house undecided voters were not.

The Greens take 13 per cent in both chambers, which is what a party polled under both question designs looks like. Labor and the Coalition each lose ground when the longer list is offered, seven points and nine points respectively, and the parties absent from the shorter list take 36 points of the upper-house vote between them. The lower house is not a less fragmented chamber than the upper house in this poll; it is a question with fewer boxes.

This matters for every figure built on the lower-house question. The 59–41 two-party result uses 2023 preference flows applied to four categories, one of which is an undifferentiated 20 per cent. In the March 2026 wave DemosAU added One Nation to the lower-house question, the party took 21 per cent, and the any-other-candidate share fell from 20 to 7. Read in sequence those two waves describe a question that changed, not an electorate that moved by thirteen points in four months. Victoria's October 2025 wave ran the same four-option question and is read the same way in ACE-IU-2026-020.

02

Both major parties go backwards in the upper house, and five seats are in doubt

Labor is down 6.6 points on its 2023 upper-house vote and the Coalition down 8.8. One Nation is up 9.1 from a standing start, and Family First up 3.7.

Each bar is the 95 per cent interval on a party's upper-house first preference, from the effective sample of 608, and the dot is the share itself.

Eleven further parties and independents take 1 to 3 per cent each and account for the remaining 16 points. The ordering at the top is established: Labor's nine-point lead over the Coalition clears the roughly ±5.6 points a within-poll gap of that size carries on this sample, and the Coalition's six points over One Nation clear about ±4.7. One Nation's two points over the Greens do not, against about ±4.2, so the poll places those two in no order.

DemosAU converts these shares into a seat estimate for the twenty-one members elected at each New South Wales election, carried alongside the twenty-one continuing members elected in 2024. It projects three seats for One Nation, which holds none of the twenty-one continuing seats, and leaves five of the twenty-one in doubt.

Both columns sum to 21 and the chamber to 42, which is the size of the Legislative Council. The projection is DemosAU's, made by its head of research from these shares; it is not an Ace Strategies estimate and it carries the uncertainty of the shares it is built on.

A Labor government holding 14 of 42 would need eight further votes to pass legislation, and the Greens' four would not supply them. That is the point DemosAU's own note draws from the table, and it is the structural finding of the wave: the chamber the poll projects is one in which the government's working majority has to be assembled from more than one minor party.

03

Twenty points of two-party vote separate voters in their forties from voters over 55

DemosAU publishes no cohort sample sizes for this wave. If the three age groups each hold about a third of the effective sample of 608, a two-party gap between two of them carries roughly ±9.7 points, and gender halves carry about ±7.7.

Every first-preference row sums to 100 in the report. The undecided column sits outside that total and is excluded from the party shares. Cohort figures rest on subsamples and carry wider margins than the ±4.0% statewide topline.

Age is the division this poll measures most clearly. Labor takes 70 per cent of the two-party vote among those aged 35 to 54 and 50 per cent among those 55 and over, a twenty-point gap against an interval of about ±9.7. The mechanism is visible in the primary columns: in the middle group Labor takes 45 and the Coalition 20, a 25-point lead, while among the over-55s the Coalition's 39 edges Labor's 36 and the two-party result is a dead heat. The Greens run from 26 per cent of under-35s to 4 per cent of over-55s, a 22-point gap against about ±6.9, and that 4 is the lowest figure any party records in any cohort here.

Two further readings are worth separating from the noise. Renters give Labor 68 per cent after preferences and outright owners 52, a sixteen-point gap that clears its interval, and renters also hold the highest undecided share in the table at 15 per cent, so that group is both the most favourable to Labor and the least settled. The seven-point gap between men and women on the Labor primary, 41 against 34, sits inside about ±7.7 and should not be read as a difference.

04

October against the March 2023 election

Percentage points, lower-house first preferences. Labor's primary vote is unchanged and the entire two-party movement comes from the Coalition's fall.

The changes are DemosAU's own. The any-other-candidate movement reads +2.1 against a difference of 3.8 in the columns beside it because its baseline is not the 16.2 others took in 2023: a change of +2.1 implies 17.9, which is the others and One Nation combined, at 16.2 and 1.8. That is the consistent comparator for a question with no One Nation box, and it is a second place the omission shows up. The two-party baseline of 54.3 is likewise the figure DemosAU's stated change of +4.7 implies.

Labor's position two and a half years into the term is its election result exactly, and the two-party gain of 4.7 points is carried entirely by the Coalition's 5.4-point fall. Nothing in this wave shows the government adding support; it shows the opposition losing it, with the Greens and the unnamed others collecting the difference. The question of where the Coalition's lost votes sit is one this ballot cannot answer, because the only destination it offers is a category rather than a party. The March 2026 wave answers it, and ACE-IU-2026-025 reads that answer.