Tasmania, September 2026

01

No party has moved more than two points since June/July

The table sets first preferences at the July 2025 election beside DemosAU's four Tasmanian waves since, and One Nation was first offered to respondents in June/July 2026.

The bars show first preferences at the July 2025 election and in the June/July and September 2026 polls, in per cent. The chart combines Shooters, Fishers and Farmers with Others, and the table separates them.

The Liberal vote fell in two steps between October 2025 and June/July 2026, six points to January/February and seven to June/July, and has held at 28 since. Labor has stayed between 21 and 24 across all four waves and the Greens between 14 and 15. Independents have fallen from 17 in January/February to 12 and then 10.

The lines trace first preferences from the July 2025 election to September 2026, with each poll placed at the middle of its fieldwork. The waves sampled 1,021 (October 2025), 1,071 (January/February 2026), 999 (June/July 2026) and 1,097 (September 2026), and the open circle on each line is the election count.

Figures are first preferences in per cent. The polls exclude undecided voters (3% in September, 4% in June/July), and One Nation was not offered before June/July 2026.

02

The Liberal lead is outside sampling error and second place between Labor and One Nation is open

The effective sample of 761 from 1,097 respondents (a design effect of 1.44) gives ±3.6 points on a 50 per cent share. Each party's interval is narrower because its share is smaller. The intervals cover sampling error only.

Each bar is the 95 per cent interval on a party's September share, from the effective sample of 761, and the dot is the share itself.

Each gap is between two shares of the same sample, and its interval uses the multinomial variance.

The Liberal lead over both Labor (6 points) and One Nation (7 points) lies outside sampling error. The 1-point gap between Labor and One Nation lies inside it, so the two sit level in second place.

Each change is between two independent samples, with effective sizes of 732 and 761.

All seven changes since June/July lie inside their intervals. The largest, independents at −2, carries an interval of ±3.2, and the two waves are consistent with no change in any party's support.

Against the 2025 election, which is a count with no sampling error, the Liberal fall of 11.9, the independent fall of 5.3, the Labor fall of 3.9 and One Nation's 21.0 each exceed their intervals. The Greens' +0.6 lies inside its interval of ±2.5.

03

One Nation arrived on 21 as the Liberals, independents and minor parties fell 18 points, and Labor and the Greens 3

DemosAU first offered One Nation to Tasmanian respondents in June/July 2026. The change from the last wave without the party (January/February) to the first wave with it (June/July) shows which categories fell as it arrived. The waves are five months apart and the comparison is between totals, so every figure is a net change. DemosAU publishes no individual-level movement.

The bars show the change in each party's first preference vote from January/February to June/July 2026, in points.

Figures are first preferences in per cent, and the change intervals use the effective samples of 745 and 732.

The six categories that fell lost 21 points between them, the same as One Nation's share. The Liberals account for a third of those points (7), independents a quarter (5), Shooters, Fishers and Farmers and Others 29 per cent together (6), Labor 10 per cent (2) and the Greens 5 per cent (1). The Liberal, independent, Others and Shooters, Fishers and Farmers falls each exceed their intervals, and the Labor and Greens falls lie inside theirs. A net change can hide movement in both directions: a party can lose voters to One Nation and gain others elsewhere, and the poll cannot show it.

The bars show the change in each party's first preference vote from the July 2025 election to September 2026, in points.

Figures are in per cent. The four parties that lost vote lost 22.0 points between them, and One Nation, the Greens and Others gained 22.0.

One Nation's 21.0 points equal 95 per cent of the 22.0 points lost by the four parties whose vote fell. The Liberals account for 54 per cent of those points, independents 24, Labor 18 and Shooters, Fishers and Farmers 4.

One Nation polls between 21 and 23 in the four latest DemosAU state polls

Each figure is One Nation's first preference in the latest DemosAU wave in that state, in per cent. Sources: the Ace Strategies poll briefings ACE-IU-2026-015 to 017 and the Victorian poll register.

The four readings sit in a two-point band from 21 to 23. Tasmania's 21 is level with Western Australia's for the lowest, in a state where the party has no elected representatives.

04

EMRS's August poll matches DemosAU on One Nation and on the order of the five parties

EMRS polled 1,000 Tasmanians on 17 and 18 August, before One Nation's registration as a Tasmanian party, with a stated margin of error of ±3.1 per cent. DemosAU's June/July and September waves both place One Nation at 21, and so does EMRS.

Each bar is DemosAU's September 95 per cent interval from the effective sample of 761, the dot is DemosAU's share and the diamond is the EMRS August reading.

DemosAU 95% interval DemosAU September EMRS August

Figures are in per cent, with undecided voters excluded in both polls. EMRS reports five categories that sum to 100, so its independent figure carries every vote outside the four named parties.

EMRS lies inside the DemosAU interval for four of the five named parties. The exception is independents (EMRS 14 against an interval of 7.9–12.1), and the categories explain it. DemosAU offers Shooters, Fishers and Farmers and Any Other Candidate as separate options at 2 per cent each, and EMRS reports neither. DemosAU's independents, Shooters, Fishers and Farmers and others total 14, the same as the EMRS independent figure.

Figures are in per cent, and the EMRS undecided figure is the remainder of 100. The two houses differ in the size of the undecided group and agree on Rockliff's share among those who name a leader.

Both houses put Rockliff ahead by 7 to 8 points and at 55 per cent of respondents who name a leader. EMRS's lead fell from 19 points in May to 7 in August. DemosAU's moved from 9 in June/July to 8 in September, within sampling error.

05

Liberal and One Nation voters supply four in five of Rockliff's points, and Labor and Greens voters seven in ten of Willie's

Rockliff leads Willie 42–34 as preferred premier, against 41–32 in June/July. The undecided share fell three points to 24, inside the interval of ±4.4 on that change. The lead of 8 points carries an interval of ±6.2, so its lower bound is +1.8.

Each bar is a leader's preferred-premier share of all respondents, divided by the voting intention of the respondents who named that leader. The division is derived from the published crosstab and the topline, and the tinted remainder of each track is everyone who named the other leader or neither.

Each cell is a group's preferred-premier percentage multiplied by its share of respondents, using September voting intention with 3% undecided. Rounding error is up to ±0.5 points per row.

One in five of Rockliff's points (9.0 of 42) comes from One Nation voters, and one in five of Willie's (7.3 of 34) comes from Greens voters. Liberal voters supply 60 per cent of Rockliff's points and Labor voters 51 per cent of Willie's. One Nation voters also supply 4.7 points to Willie, 14 per cent of his total.

Each row is one voting-intention group's preferred-premier split, in per cent of that group. † The report does not label this share, and it is shown as the remainder of the row.

Among One Nation voters 44 per cent name Rockliff, 23 name Willie and 33 are undecided; in June/July the split was 42, 19 and 39. Labor voters name Willie at 82 per cent, with 14 undecided and 4 naming Rockliff. Liberal voters name Rockliff at 93 per cent. The undecided share fell most among One Nation voters (39 to 33), independent voters (52 to 41) and voters with no party choice (76 to 71). These are subgroup readings: One Nation voters are about 160 of the effective sample, which puts an interval of about ±11 points on the change from 39 to 33, and the other two groups are smaller.

Rockliff rates highest of the party figures on a net of 0, and the two treasurers lowest

Each bar is a figure's September net rating, the positive share minus the negative share, drawn left for a net negative and right for a net positive.

Figures are in per cent. The change column follows the June/July net rating in every row where June/July tested the figure; Archer, Haddad and O'Byrne were not tested then.

• David O'Byrne (+9) is the only figure whose net rating is positive outside sampling error (±4.6). Kristie Johnston's +4 lies inside the same interval. Independents as a voting option fell 5.3 points from the 2025 election to 10 per cent.
• Jeremy Rockliff's net of 0 is the highest among the party figures. His positive share (35) is the highest on the list and his neutral share (30) is the lowest.
• The Treasurer (−18) and the Shadow Treasurer (−14) hold the two lowest nets of the ten.
• The three Liberal figures measured in both waves moved up (Rockliff +3, Barnett +6, Abetz +7), Willie fell 2, Winter rose 2 and Woodruff is unchanged. Every one of these changes lies inside its interval, which on a change in net rating between two waves of this size is ±6.6 to ±8.5 points.
• Neutral is the only middle answer and the question offers no separate don't-know, so neutral includes respondents who do not know the figure. It exceeds half for Johnston (58), O'Byrne (57) and Haddad (57).
• September tests Archer, Haddad and O'Byrne and drops Peter George and Craig Garland, who were tested in June/July.

06

One Nation polls 28 among TAFE-qualified voters and 16 among the university-educated, a gap wider than subgroup error

The table sets September cohort results beside their June/July counterparts. DemosAU publishes no cohort sample sizes. If gender halves each hold half the effective sample (about 380), One Nation's 21 carries ±4.1; if age or education groups each hold a third (about 254), it carries ±5.0.

Each row marks One Nation's first preference in a cohort in June/July (open circle) and September (filled circle), in per cent.

June/July 2026 September 2026

Figures are September 2026 first preferences by cohort, in per cent. The September report also labels three small shares: Shooters, Fishers and Farmers 5 among 18–34s and 3 among TAFE-qualified voters, and Others 3 among school-educated voters.

• TAFE-qualified voters are the highest One Nation cohort in both waves (28 and 28). University-educated voters (11 and 16) and 18–34s (15 and 16) are the lowest in both. On cohorts of a third of the sample, the 12-point gap between TAFE-qualified and university-educated voters carries an interval of ±7.1.
• The Liberals are first or equal first in eight of the nine rows, and One Nation is first among TAFE-qualified voters. Every one of those leads is smaller than its subgroup interval: One Nation's 5-point lead over the Liberals among TAFE-qualified voters carries ±8.8.
• The Liberals lead 18–34s on 29, with the Greens second on 23 and Labor third on 17. Labor and the Greens are level on 17 among TAFE-qualified voters.
• Men and women are within two points on every party. In June/July the Liberal share was 26 among men and 31 among women. One Nation among women was 20 in June/July and 21 in September.
• Cohort cells moved 1.9 points on average between the two waves. The largest move is 6, independents among 35–54s (14 to 8).

07

Cost of living has overtaken housing as the biggest issue and health has risen four points

Respondents named the biggest issue in their own words, and DemosAU coded the answers to categories with the assistance of AI. Shares sum to 117 per cent (120 in June/July) because a response can fall in more than one category.

Each bar is the share of respondents naming a category as the biggest issue facing Tasmania in September, and the tick marks June/July where the category kept its name.

September 2026, first figure June/July 2026, second figure

Figures are in per cent, and changes are shown only for categories with identical names in both waves.

• Cost of living at 29 (up 5) and health at 18 (up 4) each exceed their change intervals, by 0.5 and 0.3 points. Both margins are smaller than the rounding in the published whole-number figures, which can move a change by up to a point, so both rises sit at the edge of sampling error. Housing at 23 (down 2) lies inside its interval of ±4.3.
• Cost of living, housing and health together account for 70 mentions per 100 respondents.
• Stadium spending holds at 7 per cent, level with economy and jobs.
• Crime is named by 5 per cent in Tasmania. In DemosAU's June Victorian wave the figure was 29.
• Technology and AI data centres enter as a category at 1 per cent.

08

At these shares no pair of parties reaches 18 of 35 seats

DemosAU publishes no seat projection for this wave. The benchmark multiplies each statewide share by the 35 seats of the House of Assembly, a projection from the poll's statewide shares. A majority is 18 seats, which is 51.4 per cent of the vote on this arithmetic.

The upper bar is the House of Assembly today and the lower bar is each party's September share multiplied by 35 seats, with the 18-seat majority marked. Independents today are six: five elected in 2025 and one former Greens member.

Seat-equivalents are share × 35. The Greens hold 4 seats after a member resigned to sit as an independent in January 2026.

The 18-seat line is 51.4 per cent of the vote. It lies inside the intervals for the first three rows and the last, and outside the other four.

The nearest pairs are Liberal and Labor at 17.5 and Liberal and One Nation at 17.15. Labor and the Greens together reach 12.95. Adding every independent, Shooters, Fishers and Farmers and other seat brings them to 17.85, which is 0.15 short.

DemosAU describes a multi-party hung parliament for these shares. The Conversation reads the combined Liberal and One Nation 49 per cent as a majority of seats. The benchmark puts the pair at 17.15, 0.85 below 18, and the interval on the pair's share (45.4–52.6 per cent) runs from 15.9 to 18.4 seat-equivalents, so 18 lies inside it.

The table calibrates the benchmark against the 2025 election under Hare-Clark. Differences are taken from the unrounded count (the Greens polled 14.44), so a row can differ by 0.1 from its rounded columns. Statewide vote and seat shares differed by at most 2.7 points (Labor).

The benchmark leaves out electorates, preferences and candidates

• Hare-Clark operates in five seven-member electorates with a quota of 12.5 per cent. At 21 per cent One Nation holds 1.7 quotas on an even spread, and DemosAU publishes no electorate cut, so the benchmark cannot show where the party's vote is concentrated.
• The poll carries no preference data, so the benchmark cannot show how One Nation, Liberal and independent preferences would move between candidates.
• One Nation has no candidates, and the generic independent option cannot separate sitting crossbenchers from other candidates.
• The benchmark converts every share to seats at the same rate.

09

Three dates follow the poll

The next Tasmanian House of Assembly election is expected in 2029. DemosAU has announced no date for its next Tasmanian wave.

Sources: the Tasmanian Electoral Commission timetable and published reports.