Tasmania, February 2026

01

Six points come off the Liberals and no party takes them

The Liberal vote falls from 41 to 35, against an interval on that change of about ±5.5. The seven points the two major parties lose between them turn up exactly in the four other columns, and not one of those gains is large enough to measure on its own.

The February figures exclude 3 per cent of voters who were undecided and the October figures exclude 2 per cent. The change intervals use the two effective samples, 500 in October and 745 here, and each party's own share. The two-point rise for the Shooters, Fishers and Farmers sits just outside an interval of ±1.9, but it rests on about ten respondents in October and thirty here, which is too few for that interval to be relied on. No two-party preferred figure is published; the report records the calculation method as not applicable.

Among the four parties with enough support to measure, the Liberal fall is the only change since October that clears its interval. Labor's single point, the independents' three and the Greens' zero all sit inside theirs, and a reader should treat those three as consistent with no movement at all rather than as small movements in a direction.

Where the six points went is a separate question from whether any one destination can be seen. The Liberals and Labor lose seven points between them and the other four columns gain seven: independents three, the Shooters two, the residual category two and the Greens none. That is the arithmetic of a loss that disperses. A six-point fall split three or four ways produces three or four gains each too small to distinguish from noise, even though the sum of them is certain because the shares must total 100. The combined major-party vote is the figure that carries the finding: 65.8 at the July election, 65 in October, 58 here.

So the wave records a government losing support without an opposition collecting it. Labor's 23 per cent is a point below its October figure and 2.9 below the 25.9 it took at the election, and on every cohort in section 03 it sits between 20 and 26. Three months after a poll that found nothing moving at all, the thing that moves is the Liberal vote, and it moves outward rather than across.

02

The poll ranks the first two places and declines to award the third

Twelve points separate the Liberals from Labor and six separate Labor from the independents, and both gaps clear their intervals. Two points separate the independents from the Greens against an interval of about ±4.1, so on this evidence third place is unresolved.

Each bar is the 95 per cent interval on a party's February first preference, from the effective sample of 745, and the dot is the share itself.

Labor's interval runs from 20.0 to 26.0 and the independents' from 14.3 to 19.7, and the two do not meet. The independents' interval and the Greens' run from 14.3 to 19.7 and from 12.4 to 17.6, overlapping across more than three points. The right test on two shares inside one poll is not whether their intervals overlap, because that test is too conservative and will call differences unresolved that are not; it is the interval on the gap itself, which here is ±4.1 against a gap of two. Both readings point the same way in this case, and the finding is that the poll puts the independents and the Greens in no order.

That matters more in Tasmania than the same result would elsewhere. The House of Assembly returns five members in each of seven divisions, thirty-five seats in all, and a quota in a five-member division is one-sixth of the formal vote, or 16.7 per cent. Taken statewide, 35 per cent is a little over two quotas, 23 is a little under one and a half, and 15 is nine-tenths of one. The independents' 17 does not convert the same way: it is the total for candidates who stand separately in different divisions and whose votes do not pool, so it buys no quota anywhere by itself. This is statewide arithmetic rather than a seat projection, the vote is not uniform across the seven divisions, and DemosAU publishes no Tasmanian seat estimate.

03

The Greens' vote has a shape and nobody else's does

DemosAU publishes no cohort sample sizes for this wave. If the three-way splits each hold about a third of the effective sample of 745, a gap between two of them carries roughly ±6.4 points.

Every row sums to 100 in the report. The undecided share of 3 per cent sits outside these totals. These figures are read from the charts on the report's pages 6 and 7, which carry no text layer. Cohort figures rest on subsamples and carry wider margins than the ±3.6% statewide topline.

The Greens take 19 per cent of under-35s and 11 of the over-55s, and 10 per cent of the school-educated against 21 of the university-educated. Both gradients run in one direction without a reversal in the middle, and both clear the roughly ±6.4 points a three-way split of this sample supports. They are the only cohort differences in the table that do.

Everything else is flat. The Liberal vote runs 38 among men and 32 among women against an interval of about ±6.9, 35, 34 and 35 across the three age groups, and 36, 32 and 37 across the three education groups. October produced 40, 42 and 41 on education, so this is the second Tasmanian wave in a row in which education does not divide the Liberal vote, and it is a notable absence beside Queensland, where the LNP runs from 30 per cent among the school-educated to 42 among the university-educated in its February wave. Labor sits between 20 and 26 in every cohort and the independents between 15 and 18, so neither has a group it owns and neither has one it has lost.

Two cautions belong with this table. The Shooters, Fishers and Farmers and the residual column both show gradients, falling from 7 and 7 among the school-educated to 1 and 1 among the university-educated, but at those shares a cohort cell holds only a handful of respondents and the pattern should not be read as a measurement. And the report does not break February's figures down by housing tenure, where October found a twenty-eight-point Liberal gap between mortgage holders and renters: the one division that survived its intervals in the previous wave is the one this wave does not measure.

04

One rating in nine improved, and the premier's lead did not move

Jeremy Rockliff's net is 2 points positive and the other eight run from 2 to 23 negative. Dean Winter's ten-point improvement is the only change since October large enough to separate from sampling noise.

Each bar is the net rating, the positive share less the negative share, in points. The note gives the three underlying shares and the change in the net since October, which the report publishes itself.

Dean Winter's net goes from 33 points negative to 23, and a change of that kind carries an interval of about ±8.2 across these two samples, so it clears, though not by a wide margin. The composition matters more than the headline. His positive share moves from 15 to 17, which is inside its own interval, and his negative share falls from 48 to 40, which is not. Eight points moved out of holding a negative opinion of the shadow treasurer and into holding none, and his neutral share rises from 37 to 43 accordingly. He is better thought of as less disliked than as better liked.

Rockliff is again the only figure in the set with more positive opinion than negative, at 37 against 35, and again the one fewest respondents are neutral about, at 28 per cent. He leads Josh Willie as preferred premier 43 to 32, against 46 to 34 in October. The lead is eleven points where it was twelve, which is no change. What moves is the share naming neither, from 20 to 25, and at an interval of about ±4.7 that is the only preferred-premier movement in the wave that clears. So the Liberal primary vote falls six points while the premier's standing against the opposition leader holds, and five per cent of the electorate stops choosing between them.

The three independent members tested rate at 2, 8 and 12 points negative on neutral shares between 52 and 56, and all three are a little worse than in October, when they stood at 0, 2 and 10. Peter George's six-point fall is the largest in the set after Winter's rise and is inside its interval of about ±8.0. It is worth holding those three ratings apart from the independent vote in section 01, which rose three points over the same months: that 17 per cent is spread across candidates standing in seven divisions, and most of them are not among the people this question tested.

05

Six months on from the election, on a ballot that still had no One Nation line

Both major parties sit below their July 2025 result, the Liberals by 4.9 points and Labor by 2.9, and the any-other-candidate category is 4.4 points above its 1.6. The question listed six options and One Nation was not one of them.

The bars show each party's February first preference against its July 2025 election result, in points.

A comparison against an election uses only the poll's own interval, because the election figure is a count rather than a sample. On that basis two of the six movements clear: the Liberals' 4.9-point fall against an interval of ±3.4, and the any-other-candidate category's 4.4-point rise against ±1.7. Labor's 2.9-point fall sits just inside its ±3.0, and the independents, the Greens and the Shooters are all well inside theirs. Six months on from an election that returned no majority, the measurable change is that the governing party is smaller and the category with no name on it is nearly four times what it was.

6% The any-other-candidate category took 6 per cent in February, against 1.6 at the July 2025 election and 4 in October. In June/July 2026, with a One Nation box on the ballot for the first time, it fell back to 2.

The question offered the Liberals, Labor, independents, the Greens, the Shooters, Fishers and Farmers and any other candidate, and this was the second and last Tasmanian wave to run it without a One Nation line. When DemosAU added the party in June/July 2026 it recorded 21 per cent, and the Liberals fell from 35 to 28 in the same wave. Every jurisdiction in this series shows the same sequence: a late-2025 ballot without the party, then a figure between 17 and 21 per cent at its first appearance, in Victoria, New South Wales, South Australia and Western Australia as well as here.

What that does not establish is that One Nation's Tasmanian support was waiting in the any-other-candidate column. Six points could not hold twenty-one, and the net changes ACE-IU-2026-019 sets out put 7 of those 21 points against the Liberal column, 5 against independents, 4 against the residual, 2 each against Labor and the Shooters and 1 against the Greens. Those are movements between totals five months apart rather than observed switches by individual voters, and DemosAU publishes nothing at the individual level. The reading they support is that One Nation's February support was distributed across every box on this ballot, most of it among people who told this poll they would vote Liberal.