What to watch on referendum night


When the results start flowing in on Saturday night, it will be our first experience of a federal referendum night in 24 years. For this post, I wanted to discuss some ideas of what to expect on the night, and how to assess the results as they come in.

The big difference compared to a typical election is the lack of a previous election to compare results. A typical election night prediction system relies on matching booth results to the equivalent results from the previous election. If you compare results so far specifically to the results from those same areas at the previous election, you can isolate the problem of an early sample not being representative of the whole electorate.

There was no previous referendum on the Voice, so we don’t have that point of comparison. It means we’ll need to be a bit more cautious in examining the results that come in before calling a result. But there’s still some things we can say.

Let’s start by talking about timing of results.

Votes should be quicker to count than at a federal election. There’s no preferences, just a Yes or a No. There’s no second ballot to separate. Overall this should mean that a booth of a given size should come in more quickly than at a federal election.

With the start of daylight saving, there are five main timezones to consider when it comes to results coming in. Polls will close at 6pm AEDT in New South Wales, Victoria, Tasmania and the ACT. They’ll be followed up by South Australia at 6:30pm and Queensland at 7pm. The Northern Territory will follow at 7:30pm and finally Western Australia at 9pm. It seems quite plausible that the referendum will be called by the time WA closes.

(Incidentally, polls close in New Zealand at 7pm local time, or 5pm AEDT. New Zealand results should come in very quickly, so we may know the result before referendum results really start flowing in.)

Ordinary referendum day results will come in first, and the smallest booths tend to be more conservative. We’d want to wait for a much more representative share of those booths before beginning to draw conclusions about a particular state or electorate.

Recent Australian elections have settled on a predictable pattern. There’s often a quiet period with no results once all of the ordinary booths are counted, before the pre-poll booths come in later that night.

With the referendum held over the entire country in multiple timezones, it’s unlikely there will be a national quiet period – ordinary booths from further west will likely come in while the south-east is in that quiet period.

It looks like the number of pre-poll and postal votes is on track to be more than in 2022. These vote categories have recently tended to be more conservative than the election day vote.

So it seems likely that the results in a state or electorate will skew towards No once the pre-poll vote reports in. For Yes to win, you would need the state to have a clear margin before the pre-poll is counted.

We don’t have booth matching, but there are other datapoints that can be used to put the results in context. Perhaps it can’t be used to predict the outcome, but it can certainly have descriptive value.

I think there is value in comparing the results to the last election’s two-party-preferred vote. In yesterday’s post, I looked at how the two-party-preferred vote (both in 1998 and 2022) compared to the Yes vote in the 1999 referendum.

But after writing that post, I realised a big difference between 1999 and 2023 is that this referendum is being proposed by a Labor government with a Coalition opposition clearly opposing. Perhaps a better comparison would be 1988. This was the last occasion when Labor proposed a referendum, and was also held mid-term.

Three of the four 1988 referendums were the biggest defeats in the history of Australian referendums, with a Yes vote of 30-33%. The fourth managed 37.6%.

When you compare the Yes vote in these referendums to the Labor two-party-preferred, the correlation is very high. While the 1999 Yes vote had a correlation of just under 0.4 compared to the Labor two-party-preferred vote at the election the year before, the four referendums in 1988 had correlations of 0.69 to 0.81.

As you can see in this chart, there is a very strong correlation between the Yes vote and the Labor 2PP.

If we see a similar trend in 2022, it could be that the referendum result could be quite strongly correlated with Labor’s two-party-preferred vote (although Western Australia will likely buck that trend).

Ultimately when trying to call the result in a particular state, it will be necessary to look for results in a diverse range of seats, each of which has a broad and representative sample of results.

If the referendum results are as decisive as some polls have suggested in recent days, it might not be that close, and the result could be called quite early. But it will certainly be interesting either way.

Liked it? Take a second to support the Tally Room on Patreon!
Become a patron at Patreon!


  1. Whilst not a referendum as such, the 2017 same sex pseudo-referendum should be useful as a relatively recent guide. In both cases votes are presented with a simple yes/no choice and the result is mostly a case of conservative vs. progressive and I suspect there will be similar demographics on the yes and no sides, just with the pendulum shifted more towards the “no” in the current referendum.

  2. @ David
    A lot of areas with large Anglo-Celtic areas voted Yes strongly for example Dunkely and the Hunter Region voted very strongly Yes for SSM but i doubt they would vote for the Voice. Similar i doubt believe Blaxland, Watson, Bennelong, Parramatta, Greenway etc will have the highest No in the country. SSM did not really correlate with Education for example much of Regional NSW had a better vote for SSM than much of Greater Sydney so progressive/conservative are often issue by issue rather than two blocks.

  3. @David @Nimalan

    You both make good arguments – to add to both:

    Not every No voter I know is a conservative – some are libertarian, some are middle of the road, heck, some would be left-leaning. Same how not every Yes voter is a leftist/progressive. Also, the SSM vote was 6 years ago, which is a good while for people’s opinions to change on almost anything.

    I myself voted informally in the SSM vote, and am voting No on Saturday. I’ve voted Liberal first preference every election, expect for 2016 (Australian Christians) and 2022 (WA Party)

  4. 3 states have to vote No for the referendum to fail and 4 states have to vote Yes to pass.

    Once we get solid results from NSW, VIC, TAS and SA, ideally from a mix of locations (regional, capital city, large and small booths), we can confirm whether or not it is a “No” victory. QLD is an hour behind AEDT and is much larger and more dispersed than SA and so counting will take longer.

    The growth of prepoll and postal votes and the sky-high enrolment rates means that counting may be slower than in 1999 and may be more unpredictable.

    I don’t recall the 1999 referendum but I believe once it was confirmed that NSW, QLD, TAS and SA were trending towards “No”, the referendum was over. VIC was line-ball on referendum night.

  5. @Votante wouldn’t NSW take the longest because it’s by far the most populous state in Australia? Or maybe WA because they’re three hours behind due to daylight savings and it is a very large state in terms of area.

    According to a recent Focal Data poll, only one seat in all of WA will vote Yes: the Division of Perth. Everywhere else will vote No, according to that poll. That poll also shows that the highest No vote will be in Queensland, with No polling over 80% in two seats: Kennedy and Maranoa (both in outback Queensland).

    In the same poll, the Yes vote led in just 22 seats nationwide (none were Coalition seats): Sydney (NSW), Grayndler (NSW), Melbourne (VIC), Canberra (ACT), Adelaide (SA), Fenner (ACT), Brisbane (QLD), Macnamara (VIC), Reid (NSW), Higgins (VIC), North Sydney (NSW), Clark (TAS), Bean (ACT), Perth (WA), Kingsford Smith (NSW), Griffith (QLD), Greenway (NSW), Wills (VIC), Cooper (VIC), Bennelong (NSW), Ryan (QLD) and Chisholm (VIC). On the other hand, the poll found that the No vote is between 70% and 80% in the following seats: Riverina (NSW), Indi (VIC), Moncrieff (QLD), Fisher (QLD), Grey (SA), Hume (NSW), Longman (QLD), Bowman (QLD), Cowper (NSW), Lyne (NSW), Calare (NSW), Durack (WA), Herbert (QLD), Leichhardt (QLD), Fadden (QLD), Forrest (WA), Groom (QLD), Gippsland (VIC), Nicholls (VIC), Canning (WA), O’Connor (WA), Capricornia (QLD), Hinkley (QLD), Flynn (QLD), Dawson (QLD), Wright (QLD) and Mallee (VIC). And again, the poll found that No vote is over 80% in two seats, both in Queensland: Kennedy and Maranoa.

    On a side note, Sky News thinks that Hume is spelt Hyume now. Weird because even though I’m a great speller I’ve never spelt it like that. But apparently Sky News thinks it’s “Hyume” and they also think Forrest is spelt “Foreest”: https://www.skynews.com.au/australia-news/new-largescale-nationwide-poll-predicts-less-than-15-per-cent-of-electorates-will-vote-yes-in-massive-blow-to-the-voice/news-story/38e2c3587dc48e9dfc8ab1115432c5f7?amp

  6. The poll also has Lingiari on track to vote No. Solomon is also on track to vote No but with a smaller margin.

    So analysing all these stats: as expected, the Gold Coast is going to have the highest No vote for a city with over 500,000 residents (there are only eight in the entire county: Sydney, Melbourne, Brisbane, Perth, Adelaide, the Gold Coast and Newcastle). All three ACT will vote Yes, both of the two NT seats will vote No, only one seat each in WA and Tasmania will vote Yes (Perth and Clark, respectively), only three seats in Queensland will vote Yes and for a capital city the highest No vote will be Perth.

  7. And while the poll says that no federal Coalition seats will vote Yes, if we assume that every polling place in every electorate has the same percentage as the entire electorate (e.g if the Yes vote is 40% in Seat A and 60% in Seat B then imagine that every polling place in Seat A has a Yes vote of 40% and every polling place in Seat B has a Yes vote of 60%) then there may be some state Liberal seats that vote Yes. For example, the Tasmanian state seat of Clark (where, like every Tasmanian state seat at the 2021 state election, the Liberals won a plurality of the vote) will vote Yes, the NSW state seats of Drummoyne, North Shore and Ryde would vote Yes, the Victorian state seat of Caulfield would vote Yes, etc.

  8. Nether portal state seats are smaller and so some will. The only 2 federal seats that are within the margin of error are Sturt and bradfield

  9. @Anton Kreitzer
    Agree while i think there will be some correlation with 1999 referendum and SSM postal survey it is a guide only and there is some on the left who will vote No as you said as well as some on the right i dont expect Mackellar to be good for YES on Saturday no the state Seat of Mornington (Vic), a lot of Warringah, Curtin will also be No despite many of these voters voting Teal recently.

  10. I would say the biggest deviation between the SSM and the voice vote is that Asian/Indian immigrants (esp post-2000) are more likely to vote yes this time.

    Polling suggests that while Warringah voted more YES on SSM than North Sydney, this would be reversed on the Voice (Warringah has NO winning, NSyd has YES winning)

    Even Bennelong and Reid (far more socially conservative than both of the earlier) are going better for YES than Warringah.

    I could rely on the republic referendum but I would agree with some that there have been a massive demographic change since then. My seat of North Sydney is one of them (many Chinese people came here after 2001)

  11. As for immigrants, I don’t think the Yes vote would do well for first generation Chinese Immigrants, Australian Born Chinese as well as 1.75 generation might do significantly better due to education but their percentage in the voting bloc is probably still too low to counter the conservative vote. The only probable reason why the No vote isn’t that high in Chinese Heavy seats is most PRC immigrants don’t have Australian citizenship due to China not allowing multiple citizenship and much of the immigrants came recently.

  12. @Nether Portal, “Wouldn’t NSW take the longest because it’s by far the most populous state in Australia? Or maybe WA because they’re three hours behind due to daylight savings and it is a very large state in terms of area.”

    The more booths there are, the more counters there should be. NSW is more decentralised than Vic or SA, theoretically there should be more small booths and they get counted faster. On the flipside, we in NSW have the highest enrolment rate in the country, though not every enrolled person will actually vote.

    What might drag out NSW and Victoria’s results are their electoral geography. There will strong Yes electorates and strong No electorates. If inner-city electorates e.g. Sydney, Grayndler, Wentworth, North Sydney, are producing weak Yes results, the Yes camp can write off the state. Conversely, if rural and regional electorates are producing weak No results, the Yes camp can still see a chance.

Comments are closed.