How meaningful are opinion polls purporting to measure support for inserting an Indigenous Voice into the Constitution? How good are political surveys in general, for that matter?
We all know that polls taken years before an election are poor predictors of voting intentions. Unforeseen events will occur, of course, but it’s also unrealistic to expect all voters to know what they’d do under the suddenly posed hypothetical: “If an election were held today…”
As the actual vote approaches, polls get better, especially when the question becomes “How will you vote on Saturday the Xth?” (These days the question is tweaked to take account of what is increasingly a weeks-long voting period.)
Things are trickier when a referendum is being discussed. Nearly everyone knows what an election is. But what proportion could immediately describe what a referendum actually is? What about words like “enshrined,” “Indigenous Voice” and the “Constitution” itself?
Way back in 1987, a Newspoll found that only 54 per cent of respondents “knew that Australia has a written Constitution.” In 1992 a Saulwick poll put the figure at 67 per cent. More recently, in 2021, the Constitutional Values Survey found a much higher 83 per cent answering affirmative to the easier statement, “had heard of the Australian Constitution.” And the proportion who could describe how the document is amended? We have no idea.
Americans are more likely to know about their founding document, and it’s possible Australians are more aware of America’s too. Its clauses (particularly its amendments) feature regularly in international news, commentary and popular culture. (America’s, unlike ours, is amended by majorities of federal and state legislatures.)
Until this year the Voice to Parliament was largely a preoccupation of the political/academic/media class. A little over half the respondents to this 2021 survey, for example, had never heard of it.
Given all these uncertainties, springing a benignly described question on an unsuspecting citizen isn’t going to produce a reliable indicator of people’s eventual voting decision. So it’s little wonder early polls on referendums have a history of being wildly wrong — much more wrong than voting intentions ones. But they too become better predictors as the day approaches.
(The 2017 marriage equality survey would also have suffered from the evolution of the question’s meaning, but to a much lesser extent. It was not intended to change the Constitution, and the “vote” was just another survey, labelled as such, voluntary, filled in at home — and by three-quarters of respondents, it turned out, during the first week of the official campaign. Like the 1999 republic referendum, it dealt with a familiar, long-discussed topic, but without the earlier one’s constitutionally ordained, and fatal, requirement that a specific model be approved.)
The last time a Labor government held a referendum — a midterm set of four questions — was in September 1988. Shadow cabinet voted to support two and oppose two, but the party rooms overturned them and the Liberal and National parties campaigned energetically against all four.
Triumphantly as it turned out; what the Hawke government saw as a set of proposals so inoffensive it would slip through unharmed became the worst-performing in referendum history. The cause was not helped by a High Court finding that some of the government’s info-ads had broken the law.
In the final six months of that campaign, polled support halved, from the high 60s and low 70s to the 30s. Party-support surveys have been known to shift by several points over similar periods, but nothing approaching 30 per cent. The actual survey questions have disappeared into the ether, and they would have changed over the months, but the early ones (to take one of the four proposals) might have been along the lines of “Do you support recognising local government in the Constitution?” To which a reasonable answer might have been “Sure, why not, it makes sense.”
By referendum day, after an all-singing, all-dancing campaign, the act of voting had become more complicated for the one-in-three voters who ended up “changing their mind.” From their point of view, the question might have become “Exactly why does this government want to change the Constitution?”
From there, the questions would have multiplied: “We’ve survived this long without this change, why do it now? This important document should not be tinkered with lightly; I read somewhere it will create a lawyers’ picnic. And the taxpayer dollars to do all this” — $30 million–plus was bandied about then; for the Voice the popular estimate is $200 million — “would have been much better spent elsewhere. And it wouldn’t hurt to remind this rather arrogant and complacent government who’s in charge.”
During 2022 and 2023 the main Voice polls have measured expressed opinions about “support/in favour” rather than voting intentions for a referendum held either “today” or later in the year. The wording will change later in the year, but these are the reported questions for recently released surveys.
Essential asks: “As you may be aware, there will be a referendum held later this year on whether a Voice to Parliament for Aboriginal and Torres Strait Islander people should be enshrined in the Constitution. Do you support an alteration to the Constitution that establishes an Aboriginal and Torres Strait Islander Voice?”
Newspoll, in the Australian: “There is a proposal to alter the Australian Constitution to establish an Aboriginal and Torres Strait Islander Voice to Parliament. Are you personally in favour or against this proposal?”
Resolve, in Nine papers: “The new federal government has committed to a referendum — a national vote — on whether to enshrine an Indigenous Voice to Parliament in the Constitution. You will be asked to vote on this change to the Constitution in the next year or two, and voting is compulsory. Given this, do you support an alteration to the Constitution that establishes an Aboriginal and Torres Strait Islander Voice?”
Resolve’s reported “next year or two” and “new federal government” look like gremlin-induced remnants of 2022 polls. Apart from that, its wording seems best because, while rather long and laborious, it doesn’t assume people know that the Constitution can only be amended by popular vote, or what a “referendum” is. But still, like the others, it’s different from a standard political poll because it doesn’t ask about voting intention.
So there’s variety in pollsters’ questions, much more than is found, again, in party-support polls. Referendum polling seems an even less exact science than general election polling. Despite that, though, the polls are all recording similar levels at the moment: around 60 per cent support once you exclude undecideds and/or push them to choose.
(That Resolve survey was taken in two portions. The first in December found 62 per cent net support, while the second in late January, after opposition leader Peter Dutton had launched his quasi-No “confusion” campaign and the topic started featuring heavily in the news, had it lower, at 58 per cent.)
Obviously the surveys taken in the final week of the campaign will more resemble each other and be very different from those above. They’ll ask people how they voted if they’ve done so already, or how they intend to vote. They’ll all be pretty close to the final result. (Even a 2019-sized poll fail will appear respectable unless the “error” happens to account for the difference between success and failure.)
Afterwards, accounts of the Voice referendum will describe a trajectory of surveyed “support,” but in reality the question respondents answer, from January to referendum day, will gradually have changed.
How meaningful are opinion polls purporting to measure support for inserting an Indigenous Voice to Parliament into the Australian Constitution? At the moment, barely meaningful at all. •