By Chas Underwood

This piece first appeared as a featured article in volume 95, issue one of Pelican. You can view our print archive here.

In my last article for Pelican written in June 2023, I analysed public opinion polling on the proposal to insert the Voice, an Indigenous representative body, into the Australian Constitution. I concluded the Yes campaign would probably lose: support was continually declining and there were no good reasons to expect this trend to change.

This was basically what happened in the lead-up to referendum day. Support for Yes, which had dropped fifteen per cent from its peak, steadily dropped a further ten per cent leading into 14 October. No’s eventual victory with 60% of the national vote was about the magnitude polling aggregations had predicted.

Why did we change our minds so rapidly?


Aggregate of polling for the 2023 referendum (Teratix, CC-BY-SA 4.0, Wikimedia Commons)

First, we must place the referendum’s failure in its proper historical context. The statistical heuristics of “eight successful referendums from forty-four attempts”, “one Labor victory from twenty-five attempts”, and “no victories without bipartisan support” were, it seemed, mechanically inserted into every referendum analysis throughout the campaign.

Less well-known was the tendency for referendums to start off with sky-high support before spectacularly plummeting across the course of the campaign proper. This fate notably befell Robert Menzies’ 1951 anti-communist referendum, which, starting from 80% support in Gallup polling, collapsed to the extent it lost its national majority.

Bob Hawke’s 1988 quadruple-header ran into similar problems: the proposals, concerning local government, parliamentary terms and civic rights, started with support in the mid-60s-to-70s before dropping to the low 30s by the crucial day; “Rights and Freedoms” remains Australia’s worst-performing referendum ever.

The 1999 republic referendum suffered a comparable fate, though not to such an extreme extent – the Tasmanian electoral analyst Dr. Kevin Bonham found support for the proposed model was simply not particularly high from the very beginning, with its initial peak only a few points over 50%. Still, the pattern of losing support from an initially promising position recurred.

For many, it therefore came as no surprise when the situation repeated itself for the Voice referendum. But why did it recur this time?


First, let’s consider the explicit reasons people gave for opposing the Voice when polled. Far and away, the leading response was some variant on opposing “division by race”, a result that was robust across many different polls and methodologies. This is perhaps no surprise, as it was the key slogan of the most prominent branch of the No campaign – Fair Australia, led by Jacinta Price. Price also enjoyed one of the highest net-popularity ratings among politicians in Resolve and Morgan polls.

However, the picture complicates when considering “soft” voters – those who rated themselves less sure in their opinion – and those who switched from Yes to No over the course of the campaign. Although division was another strong concern here, the leading reason related more to lacking details on the Voice or scepticism over whether it would be effective at all. This suggests although the No campaign itself might have been the spur for the greatest number of No voters, reluctance by the Albanese government to commit to a more detailed model might have proven more decisive, in the sense it swayed voters who might otherwise have stayed in or joined the Yes camp.

Structural factors also proved important. The more progressive a party you supported, the more likely you were to support the Voice. Both progressive and conservative commentators have highlighted the referendum’s failure to win bipartisan support as a factor in its defeat, the former blaming Dutton for withholding conservative support and the latter blaming Albanese for not doing enough to seek conservative support. Whether you believe fault lies with Dutton or Albanese for failing to generate bipartisan support basically comes down to your pre-existing views on the leaders and the Voice itself, so my analysis is unlikely to be helpful either way. For the record, I believe the Yes-supporting conservatives Julian Leeser and Frank Brennan put forward reasonable compromise wordings that might have garnered greater Liberal support had the Referendum Working Group adopted them. However, the overall picture is complicated by the fact that as many as 30–45% of Labor voters also voted No, who were obviously less likely to be swayed by Dutton’s decision. A more plausible overall explanation might be a loss of trust in the government in general, not merely among Liberal voters.

The more education you had completed, the more likely you were to vote Yes in the referendum. Some people interpret this tendency as a “gotcha” – “smart people vote Yes, morons vote No” – but I don’t think that’s justified. It’s no secret that university graduates tend to skew towards progressive politics, and the referendum was a progressive project. So rather than a dunk on No voters, the educational divide observed in the referendum is probably best interpreted as just another signal of the cultural divide between those with and without a university education.

Much ink has already been spilled arguing over the role of racial prejudice in the referendum result. It’s clear at least some participants, particularly on the internet, marred the conversation with racist diatribes. What is less clear is whether these commenters were loud minorities or genuinely representative of the broader community. Surprisingly, at the time of writing, there does not seem to have been any systematic attempt to investigate the question. The best we can do is work off inference and contextual clues. For instance, it’s difficult to argue racial prejudice was responsible for the truly decisive drop in support over the course of 2023 – why and how would 30% of Australians develop racial prejudice in the span of months?

Some commentators argued difficulties with the cost of living were the true spur for many No voters, who wished to send a harsh message to the Albanese government about its apparent priorities. This appeared to be borne out when poorer electorates tended to vote No at higher rates. However, when polls asked people directly about their referendum views, it did not appear that people experiencing greater financial stress were more likely to answer No. A recent ANU study suggested the apparent effect of income disappears and even reverses when controlling for respondents’ education level – that is, more-educated people happen to have higher incomes, but it is their education rather than their income itself that affected their referendum views.

Leave a Reply

Your email address will not be published. Required fields are marked *