Supporters of the Indigenous voice to parliament are facing an increasingly difficult path to victory in the October referendum, with models estimating that the no campaign is leading in every state.
Models also show that the proportion of people telling pollsters they are unsure how they will vote has almost halved this year.
Support for the voice has dropped by about 21 percentage points nationally over the past year, according to Guardian Australia’s poll tracker. This is similar to the declines in South Australia, New South Wales and Queensland. Victoria has seen a smaller decline – about 18 points – while in Western Australia the decline is closer to 25 points.
The declines among 35-54-year-olds, those over 55 and supporters of the Coalition or independents or minor parties is also about 20 percentage points. Support from Greens voters and 18-34-year-olds has remained relatively steady.
Measuring the exact level of support is tricky, as polls can vary in sample size and questions asked. They can also report differently and have opaque methods around weighting. Several pollsters have also changed their questions or methodology throughout the campaign.
Guardian Australia collected results from 48 polls and 10 different pollsters to create estimates for each state. Our tracker does not predict what side will ultimately prevail in the referendum, but can help us understand the current situation and the trend in sentiment.
Winning the referendum requires a double majority – a majority of voters nationally as well majorities in four of the six states.
The ACT and the Northern Territory are not counted towards the state majority. The campaigns have identified Tasmania as an important swing state, but it contains only about 2% of the national population, making it hard to get decent sample sizes. All three have been excluded from our tracker.
You can read more about our methodology and some of the limitations of poll averaging here. Our estimates are also slightly different from ones produced by election analysts such as Dr Kevin Bonham, whose latest are in the chart below.
Bonham has been tracking how the decline in the states’ support compares with the national trend.
“So the consistent pattern is that Victoria runs ahead of the other states and Queensland runs furthest behind,” he said.
“What’s starting to happen now is that in recent polling, South Australia has been slipping behind the national average as well, which means that even if there was a big swing back to [the yes campaign] overall, it might now have issues getting four states.
“When I started doing the state breakdowns, Western Australia was at about the national average. Now its about two and a half points behind. So there’s more decline in Western Australia than elsewhere.”
What has happened already?
The Albanese government has put forward the referendum question: "A Proposed Law: to alter the Constitution to recognise the First Peoples of Australia by establishing an Aboriginal and Torres Strait Islander Voice. Do you approve this proposed alteration?"
The PM also suggested three sentences be added to the constitution:
- There shall be a body, to be called the Aboriginal and Torres Strait Islander Voice.
- The Aboriginal and Torres Strait Islander Voice may make representations to the Parliament and the Executive Government of the Commonwealth on matters relating to Aboriginal and Torres Strait Islander peoples;
- The Parliament shall, subject to this Constitution, have power to make laws with respect to matters relating to the Aboriginal and Torres Strait Islander Voice, including its composition, functions, powers and procedures.
How would it work?
The voice would be able to make recommendations to the Australian parliament and government on matters relating to the social, spiritual and economic wellbeing of Aboriginal and Torres Strait Islander people.
The voice would be able to table formal advice in parliament and a parliamentary committee would consider that advice. But the voice co-design report said all elements would be non-justiciable, meaning there could not be a court challenge and no law could be invalidated based on this consultation.
How would it be structured?
The co-design report recommended the national voice have 24 members, encompassing two from each state, the Northern Territory, ACT and Torres Strait. A further five members would represent remote areas and an additional member would represent Torres Strait Islanders living on the mainland.
Members would serve four-year terms, with half the membership determined every two years.
Much of this change has come from people who were previously classified as “unsure” or “soft” yesses and nos, according to the Accent Research principal, Dr Shaun Ratcliff, who ran one of the largest polls of the voice for the group behind the Uluru statement from the heart. The number of undecided voters has declined by 11 points over the past year, according to the Guardian Australia tracker. This can be a noisy measure, as there are differing definitions of undecided voters and not all pollsters report this figure.
“When we first started running surveys on the voice, there had been very little national discussion,” Ratcliff said. “A lot of people didn’t know about the voice.”
But the proportion of people still undecided has come down as the voice has become more prominent in the news and people have seen ads or had conversations with friends, Ratcliff says.
“Salience has definitely increased and so some of the increase in the no vote is people who had been parking themselves in ‘don’t know’ because they really didn’t know much about it and so weren’t confident enough to express an opinion.
“Some of those ‘don’t knows’ have shifted to yes, but I suspect more have shifted to no.”
This is reflected in the monthly Guardian Australia Essential poll, which breaks down the strength of responses into five categories, ranging from “hard yes” to “hard no”. A similar trend can be seen from most pollsters.
“Hard no has gone from massively outnumbered by hard yes to more or less the reverse,” Ratcliff said.
“The number of soft yes and soft no voters has declined over time and clearly some of those soft yes voters have shifted to no.”
Support for the voice commanded majorities among Coalition, Labor, Greens and minor party/independent voters last year. But this has decreased for all but Greens voters.
Part of this is linked to increased salience, Ratcliff said, as many of these voters would have been soft yes but might not have considered it very much. But there is also some partisanship. “What you’ve seen is Coalition voters have slowly been drifting, or not so slowly, behind the party leadership.
“If you don’t have a strong view yourself, you go, ‘well, the people I agree with on other issues, the people I vote for, what’s their position?’ The Coalition is, particularly at the federal level, sending a pretty strong signal that they oppose the voice.”
Bonham said age is one of the best explanatory demographics, but data on voice support by age group is quite sparse. Different pollsters use different age ranges when reporting, if they break it down at all.
According to data from the monthly Essential poll, there have been similar declines in support among those aged 35-54 and 55-plus as what has been seen in the states. After spiking in the period to June, support among those aged 18-34 is currently only 1.5 points below what it was last August.
But more than half of the decline in the 35-54 and 55-plus age groups has been since June, when Essential changed their methodology to include responses from people who were unsure. Support among the 18-34 age group had actually risen before the methodology change.
Notes and methods:
To estimate trends in states and undecided voters, Guardian Australia fit a generalised additive model to the data
Poll results that included undecided or unknown voters were removed to give a yes/no response
Guardian Australia contacted pollsters to ask for sample sizes for each state and territory, age ranges and political affiliation but were unable to secure them for most polls, so the models were not weighted by sample
