New South Wales, March 2026

01

Thirteen of One Nation's twenty-one points come out of the box it used to be inside

A change between the October and March waves carries about ±5.4 points at 95 per cent, from effective samples of 608 and 728. Two of the four movements clear it, and both belong to the question change rather than to a shift in opinion.

Each bar is the change in a party's first preference from October 2025 to March 2026, in points. The pale band is the ±5.4 points a change of that size carries.

One Nation has no bar because it has no October figure on this question. The four parties shown are the four the October ballot offered.

The arithmetic closes exactly. One Nation's 21 points are the 13 the any-other-candidate category lost, the 7 the Coalition lost and the 3 Labor lost, less the 2 the Greens gained. That is what adding a box to a ballot does: it names a destination for votes that were already being recorded somewhere, and the category that had been absorbing them empties out.

So the reading is not that One Nation gained 21 points of support between October and March. It is that at least 13 of those points were in the October poll already, inside a category the October report could not break down. The 7 from the Coalition and the 3 from Labor are the part that could be a real movement, and only the Coalition's fall clears the interval on its own. A fair statement of this wave is that One Nation has a measurable lower-house vote in New South Wales for the first time, that the Coalition has lost ground to it, and that the size of the party's support relative to October cannot be settled by a comparison across a changed question.

02

The 61–39 two-party figure is built on 78 per cent of the ballot

New South Wales uses optional preferential voting, so a ballot that names one party and stops is exhausted rather than distributed. The flows DemosAU applies exhaust 22 points of the primary vote before the two-party figure is struck.

The flow rates are DemosAU's, printed in its methodology statement: Greens 59.5 per cent to Labor and 7.3 to the Coalition, One Nation 11.7 to Labor and 26.1 to the Coalition, others 24.0 to Labor and 19.6 to the Coalition, with the balance exhausting in each case. The columns here apply those rates to the primary votes and are Ace Strategies' arithmetic. Each cell is rounded to one decimal place and the totals are struck on unrounded values, so a column can differ from the sum of its printed cells by a tenth.

Labor's 47.1 and the Coalition's 30.9 sum to 78.0, and 47.1 as a share of that is 60.3 per cent against DemosAU's published 61. The difference is what rounding the primary votes to whole numbers costs, so the reproduction holds, and what it exposes is the denominator. A two-party figure of 61–39 in New South Wales does not describe a contest in which 61 per cent of voters prefer Labor to the Coalition; it describes the preference of the 78 per cent whose ballots reach one of them. One Nation's flow is the reason: 62.2 per cent of its preferences exhaust, which is 13.1 points of the state's vote going nowhere.

DemosAU attaches its own warning, that the figure should be treated with caution because of the changed One Nation vote and the optional preferential system. That caution is well placed and this section quantifies it. The exhaust rate is not a rounding concern: it is a fifth of the electorate, concentrated in the party that grew. The June wave publishes no two-party figure at all, which is a defensible response to the same problem.

03

The poll does not establish whether One Nation or the Coalition is second in the upper house

One Nation's 22 and the Coalition's 21 are one point apart, against a within-poll gap interval of about ±4.8 on the effective sample of 728.

Each bar is the 95 per cent interval on a party's upper-house first preference, from the effective sample of 728, and the dot is the share itself.

Twelve further parties and independents take 1 or 2 per cent each and account for the remaining 15 points, and the figures exclude 13 per cent of voters who were undecided. Labor's seven points over One Nation clear the roughly ±5.2 points that gap carries, so Labor is first in this chamber on the poll's own terms. Between One Nation and the Coalition there is nothing to report but a tie.

The upper-house swings are the larger story. One Nation is up 16.1 points on its 2023 upper-house vote, the Coalition down 8.8 and Labor down 7.6. Both major parties have lost ground in this chamber in every DemosAU wave of the cycle so far, and the Australian Citizens Party, which did not register in October, records 2 per cent. The chamber the October wave projected as needing more than one minor party to pass legislation is, four months later, one in which the second-largest bloc may not be a major party at all.

04

One Nation runs stronger among women than men, which is the reverse of the usual pattern

DemosAU publishes no cohort sample sizes for this wave. If the three-way splits each hold about a third of the effective sample of 728, a gap between two of them carries roughly ±7.7 points, and gender halves carry about ±6.0.

Every row sums to 100 in the report. Cohort figures rest on subsamples and carry wider margins than the ±3.6% statewide topline.

One Nation takes 25 per cent of women and 18 per cent of men. The seven-point gap sits just outside an interval of about ±6.0, which makes it a reading worth recording rather than a firm finding, and it runs against the pattern the party shows in most Australian polling. It is not a one-wave artefact: the June wave puts the same gap at eight points in the same direction, and two waves pointing the same way is a better reason to watch the series than either wave alone.

The clearer divisions are education and age. One Nation takes 26 per cent of the school-educated and 14 of the university-educated, a twelve-point gap against about ±7.7, while Labor runs 42 among the university-educated and 27 among the school-educated. The Greens take 36 per cent of under-35s and 4 per cent of over-55s, a 32-point gap and the widest cohort division in the three New South Wales waves. Labor's 44 among those earning over $125,000 against 30 among those under $45,000 is a fourteen-point gap that also clears its interval, and it is the inverse of the income pattern Labor shows federally.

05

March against October and the 2023 election

Percentage points, lower-house first preferences. The October column is from a question that did not list One Nation, so the final column compares two different ballots.

One Nation's 2023 figure of 1.8 is the baseline DemosAU uses, and it is recoverable two ways: the party's stated change of +19.2 against this wave's 21, and of +25.2 against June's 27, both imply 1.8. The October others figure of 20 is measured against a baseline of 17.9, the 2023 others and One Nation combined, because that question had no separate One Nation box.

Against the election the shape is clear enough to state. The Coalition has lost 12.4 points, One Nation has gained 19.2, Labor has lost 3.0 and the Greens have gained 5.3. Three years into the term the government's primary vote is close to where it started and the opposition's has fallen by a third, with the difference sitting with a party that barely contested the last election. The two-party column rises to 61 across that, but section 02 is the reason to hold it lightly: it is struck on a shrinking share of the ballot.