Federal, July 2026

01

Twice the sample, and the same question still has no answer

The effective sample rises from 871 to 1,532 and the margin of error narrows from ±3.3 to ±2.5. Three points separate One Nation from Labor against a gap interval of about ±3.7, so a third consecutive wave declines to order the first two places.

Each bar is the 95 per cent interval on a party's July first preference, from the effective sample of 1,532, and the dot is the share itself.

The poll excludes 7 per cent of respondents who were undecided. The change intervals use the two effective samples, 871 in June and 1,532 here, and each party's own share. The election baselines are reconstructed from the swings DemosAU states and sum to 100.0. No two-party preferred figure is published; the report records the calculation method as not applicable.

Not one of the five changes since June clears its interval, the Coalition's three-point recovery included. A reader should treat this wave as consistent with the June wave in every column rather than as a month in which the Coalition gained ground, and should be slower still to attribute the three points to any event of the intervening weeks.

The orderings are a different matter, and this is where the larger sample earns its keep. Labor's five points over the Coalition clear an interval of about ±3.4 and One Nation's eight clear about ±3.5, so the Coalition is confirmed in third for the second wave running and on firmer ground than in June. One Nation's three points over Labor sit against about ±3.7 and do not. Across three waves the gap between the two parties at the top has been two points, three points and three points, on effective samples of 754, 871 and 1,532, and on none of those occasions has the instrument been able to say which of them is in front.

Against the May 2025 election every party has moved further than this poll's own interval, the Greens included for the first time in the three waves: One Nation up 22.6 points, the Coalition down 10.8, Labor down 8.6, the Others down 5.0 and the Greens up 1.8 against an interval of about ±1.7. That is a consequence of the sample size rather than of any change in the Greens' vote, which has run between 13 and 14 per cent throughout.

02

One Nation holds 98 per cent of its own voters and 45 per cent of those outside the 2025 electorate

The report breaks this wave's vote down by how each respondent says they voted in May 2025. One Nation's retention returns to the 98 per cent it recorded in May after June's 94, and the Coalition's recovers from 51 per cent to 57.

The report draws this table as stacked bars and prints a figure only where a segment is wide enough to hold one. An en dash marks a cell the report leaves unlabelled; each of those rows balances to 100 across the cells shown plus the unlabelled remainder, and this briefing does not write a figure into a cell the report does not print.

Read across the three waves, the Coalition row moves and the One Nation row does not. Coalition retention runs 60, 51 and 57 per cent while the share of its 2025 voters choosing One Nation runs 37, 43 and 37. One Nation's own retention runs 98, 94 and 98. The volatility in this series lives entirely on one side of that exchange, which is consistent with a Coalition vote that is loosely held and a One Nation vote that is not.

The last row deserves more weight than it usually gets. Forty-five per cent of respondents who do not recall voting in 2025 or say they did not vote would vote One Nation now, against 18 per cent for Labor and 13 for the Coalition. This group carries the least reliable stated intention in any poll, because turning out is the thing its members did not do, and One Nation's share of this wave therefore rests more than any other party's on people whose attendance at an election is the open question.

03

Of seven One Nation positions, one has majority support and five are rejected

The report put seven statements associated with Pauline Hanson to the same respondents who gave her party 29 per cent of the primary vote. Reducing immigration is net 40 points positive. Abolishing SBS is net 46 points negative.

Each bar is the net agreement with a statement, the agreeing share less the disagreeing share, in points. The note gives the three underlying shares.

Six of the seven nets clear their intervals, which run between about ±3.5 and ±4.3 on this sample. Nuclear power is the exception at one point negative, and on an effective sample of 1,532 that is a measured tie rather than a figure too small to read. So the electorate's position on this list is settled in six places out of seven, and in five of those six it is settled against.

The gap between the first line and the rest is the finding. A party polling 29 per cent carries one position with 61 per cent agreement and six that run from a dead heat to 46 points under water, and the two public-broadcasting items are the furthest under. That is the profile of a vote organised around a single issue rather than a programme, and it means the party's support and its platform's support are different quantities that this wave measures separately.

The same question block carried two statements about democracy and the United States. Sixteen per cent agreed that Australia needs a prime minister like Donald Trump against 72 per cent who disagreed, a net of 56 points negative, and 63 per cent agreed that Australian democracy is something to be proud of against 14 per cent who disagreed, a net of 49 points positive. Both clear their intervals several times over.

04

Albanese beats Hanson head to head and ties with Taylor

On three choices Albanese takes 35 per cent, Hanson 26 and Taylor 22. Put against each alternative on its own, Albanese leads Hanson 42 to 37 and is level with Taylor at 38 apiece.

On the three-choice question Albanese is unchanged on June, Hanson is down two, Taylor up three and the don't-know share down one; none of those four changes clears its interval. The two head-to-head questions have no June counterpart in this report.

The two head-to-head results point in opposite directions from the three-choice question, and the reason is in the don't-know column. Against Hanson, 21 per cent will not choose; against Taylor, 24 per cent. Taylor's 22 per cent on three choices becomes 38 against Albanese alone, a rise of sixteen points, while Hanson's 26 becomes 37, a rise of eleven. The Opposition Leader gains more from the removal of the third name than the One Nation leader does, which is what one would expect of a candidate whose party's voters have somewhere else to go and whose own party's voters have nowhere.

So the poll supports two statements that sit awkwardly together, and both are inside its own error. Hanson is ahead of Taylor when all three names are offered, by four points against an interval of about ±3.5. Taylor does better than Hanson when the choice is narrowed to two. Neither of those is a prediction of anything; together they say that preferred-prime-minister figures in a three-cornered contest are an artefact of the list as much as a measure of the people on it.

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 June, which the report publishes itself.

Albanese leads both head-to-head questions or ties them while carrying the worst net rating of the three, at 23 points negative against Hanson's 9 and Taylor's 3. A reader should take the preferred-prime-minister figures as comparative and the net ratings as absolute: they are measuring different things, and the second set says that none of the three is liked.

The report also asked who would be the better prime minister for each of ten groups. Albanese leads on five and four of those margins clear their intervals: men by ten points, employees and workers by nine, capital city residents by eleven and low income earners by six. His four-point lead on families does not clear. Hanson leads on one group, country residents, by eight points, which does clear. The other four are level, and one of them is women, where Hanson's 38 and Albanese's 37 sit against an interval of about ±4.3. The conclusion sometimes drawn from that row, that Hanson is the preferred choice for women, is not one this wave supports; what it supports is that she is the preferred choice outside the capital cities.

05

Twenty-four public figures were rated and three are established above zero

Six of the twenty-four record a net positive rating and one records zero, but only three of those six are far enough from zero for this sample to be sure of the sign. Fourteen are established below it.

Every row sums to 100 in the report across these four columns and a neutral share not shown here. This is a separate question from the three leader ratings in section 04, which offer no not-heard-of option, which is why Albanese reads 22 points negative here and 23 there.

Only Jacqui Lambie, Chris Minns and David Crisafulli are far enough above zero for this sample to establish the sign of their rating, each at five points against intervals between about ±3.3 and ±3.6. The next three positives, at two points and one and one, are inside theirs. None of the three established positives is a member of the federal parliament's two major parties: one is a crossbench senator and two are state premiers, rated here by a national sample.

The not-heard-of column is what makes this table readable. Six figures are unknown to 5 per cent of the sample or fewer — Hanson and Albanese at 1 per cent, Morrison at 3, Abbott and Rudd at 4 and Turnbull at 5 — and every one of the six is net negative, between 8 and 22 points under. The three figures this sample does establish above zero are unknown to 18, 19 and 29 per cent of it. Being widely known and being positively regarded do not travel together in this table, which is a hard environment for anyone who has held national office and an unusually forgiving one for anyone who has not.

06

The vote divides on place, tenure and education, and no longer on gender

DemosAU publishes no cohort sample sizes for this wave. If a three-way split holds about a third of the effective sample of 1,532 each, a gap between two of those groups carries roughly ±5.6 points, and a gap between halves about ±4.5.

Every row sums to 100 in the report. The undecided share of 7 per cent sits outside these totals. The report also publishes a breakdown by personal income, which runs from 23 per cent Labor and 30 One Nation below $45,000 to 26 and 27 above $125,000. Cohort figures rest on subsamples and carry wider margins than the ±2.5% national topline.

One Nation takes 39 per cent in regional and rural Australia against 21 in the inner metropolitan areas, and 35 per cent of the TAFE-qualified against 21 of the university-educated. The Coalition takes 31 per cent of home owners against 14 of renters and 28 of the over-55s against 13 of the under-35s. Labor takes 37 per cent of respondents who speak a language other than English at home against 24 of those who speak English. Every one of those gaps clears the interval a split of this sample supports, and on the larger sample they clear it by more than June's did.

The gender row no longer divides. Labor is on 27 per cent among men and 24 among women, and One Nation on 30 and 28, both inside the roughly ±4.5 points a two-way split supports. A month earlier Labor's figures were 32 and 21, an eleven-point gap that did clear, so the gender division this series recorded in June is not present in July. The only gender gap that survives here is the Greens' five points, 16 per cent among women against 11 among men.