KEY POINTS
- Australians will vote on an Indigenous Voice to Parliament on 14 October.
- An expert who predicted Brexit is damning about the Yes camp's chances.
- Another expert explains why polls produce different results, and how to read them.
A referendum expert who predicted the shock Brexit result says his model is 95 per cent sure Australians will reject an Indigenous Voice to Parliament.
Australians will , in the country's first referendum in a generation. Polling suggests support is steadily declining.
ANU political scientist Matt Qvortrup first tipped the Voice to fail without support from the Coalition in November, when most polls had support above 60 per cent.
While accepting campaigns are a “moving feast”, Qvortrup said his model - which also accurately predicted the 2014 Scottish independence result - predicts a victory for the No camp by between 3 and 6 per cent.
Qvortrup's model works by giving governments a baseline of support of 56 per cent - based on analysis of more than 600 referendums - but then removing six to seven per cent in countries with compulsory voting. He also believes support for a referendum declines each year a government has been in power.
He has found bipartisan support generally increases the chances of a referendum succeeding, while other factors - like the wording of a question, and the economic context - are also taken into account.
The Yes camp insists it can sway a large number of Australians who are either unsure or not firmly in the No camp before 14 October.
So what do the polls say, why are they slightly different, and which way are the undecideds likely to vote?
What do the polls say, and why?
Polling paints a dire picture for the Yes camp.
The polls show support for the Voice is on a downward trajectory, and struggling to get above 50 per cent in the states it needs to win. To succeed, a referendum needs an overall majority and a majority in at least four states.
Last week's Newspoll had the No vote rising to an overall majority - 53 per cent - for the first time, with support for the Voice dipping to just 38 per cent. Nine per cent of respondents were unsure.
And Resolve Political Monitor's survey, of 1,604 eligible voters, showed 43 per cent support for the Voice, and that Tasmania is the only state left backing Yes.
Polling came under scrutiny in 2019 when many major pollsters predicted a Labor win at the federal election. The Coalition recorded a two-party preferred vote of 51.53 per cent to Labor's 48.47 per cent.
Polling shows the Yes vote on a downward trajectory.
And in further bad news for the Yes camp, it found significantly more respondents considered themselves a 'hard No' (41 per cent) than a 'hard Yes' (30 per cent).
Qvortrup said that trend is typical of referendums held on "abstract" concepts.
Voters are more willing to upend the status quo when an issue is already clear in their mind, like same-sex marriage, he said.
But on concepts they haven't thought deeply about before, or don't fully understand, initial support tends to fall away when questions about the "concrete proposal” are raised.
"The vote against the status quo is less likely when it's a completely new issue, where people have to educate themselves, which they will find tedious and tiresome and [think]: why do I have to look this up?" Qvortrup said.
Support for the Voice has steadily dropped as the No camp has on how the Voice would function.
Anthony Albanese has confirmed the referendum date as 14 October. Source: AAP / Lukas Coch
Which way will the undecideds vote?
Polling companies tend to assume about two-thirds of undecided voters lean towards voting No.
Qvortrup stresses each referendum has its "own peculiarities", and it's impossible to predict exactly which way the undecideds will go in this case. But he's "quite happy" with the assumption being applied to the Voice vote.
"You know what you’ve got, you don't know what you will get," he said.
"That means that most people who should be in favour of change will be in favour of it in a lukewarm sort of way. Whereas those who know what they've got will fight ferociously tooth and nail to maintain the status quo."
Qvortrup's model predicted the shock 2016 Brexit referendum result. Source: Getty / Michael Kappeler
"They tend to be slightly more sceptical as well. So we can be relatively sure, at this stage in the campaign ... that undecided voters are more likely to vote No," he said.
Why do polls throw up different results?
Shaun Ratcliff, principal at Accent Research, said polls producing different results is not just to be expected, it’s actually desirable.
"If everyone's giving you the exact same result, down to within a percentage point of each other, that would actually be a bit more suspicious," he said.
That’s because each poll asks slightly different questions, provides people with different options to respond with, and accounts for blind spots differently.
All that can impact an outcome.
"With high-quality surveys, I wouldn't anticipate big differences. They should only create small differences in the results, but they could easily account for 1 or 2 per cent," Ratcliff said.
He also expects more accuracy now pollsters in October.
"Before we knew the final question, it was a little bit of a guessing game on what precisely [Australians will be asked]," he said.
"We had some ideas, so you could (be) reasonably accurate. But it wasn't exact ... If I asked the question one way, and then I asked you the exact referendum wording, that could drive small differences."
Polls have shown the No camp on track for victory, despite strong support for the Voice a year ago. Source: AAP / Richard Wainwright
How are polls actually conducted?
Mainly through online surveys, with respondents often given small inducements - things like frequent flyer points or donations to charity.
They can answer questions at a convenient time - a third are now completed using a mobile device - meaning they can be longer and provide a more nuanced picture.
"The beauty is that it's not as disruptive to people. You're not getting a phone call in the middle of work, the middle of dinner, or when you're out with friends," Ratcliff said.
"You can do the survey when it's convenient. The benefit of that is you're more likely to get people actually do [it]."
That can't be said for robocalls which, while cheap and able to be conducted at high volume, struggle to get responses.
"It's really easy for people to screen calls. If you don't recognise the number, you may just not answer it because you assume it's someone trying to sell you something or scam you," Ratcliff says.
Robocall surveys also produce skewed samples, which don’t reflect the voting public; young men are least likely to complete them, meaning the results are dominated by older people and women.
While pollsters can attempt to correct after the data is accrued - using ‘weights’ - it’s an imperfect science.
"A good pollster will rely on weights as little as possible, and rely on a good sample as much as possible," Ratcliff said.
"What we all try and achieve is actually finding a sample in the first place that looks like the population."
Some phone polls are conducted with a human interviewer and, if they’re skilled, produce accurate results. But they’re often too expensive to conduct at scale.
How should I read polls?
Ratcliff advises Australians to look for how transparent a company is about the question they asked, the medium they used, and their sample size.
And although methods change over time, a company’s track record of accuracy is also a factor, he said.
But he warns no individual poll can be expected to be “right on”, and looking at trends across different polls is a better way of understanding what’s happening.
"It does look like support (for the Voice) has declined over the year, and that is something we're seeing across pretty much every survey … everyone's picking it up."
Correction/clarification: A previous version of this article said a referendum required three states to succeed. This has been corrected/clarified.
Stay informed on the 2023 Indigenous Voice to Parliament referendum from across the SBS Network, including First Nations perspectives through NITV.
Visit the to access articles, videos and podcasts in over 60 languages, or stream the latest news and analysis, docos and entertainment for free, at the
