A few Regional Tables on The Voice Referendum Results

The older I get the harder I find it to pull my brain together and write analysis pieces the day after an election broadcast.

I thought the option that required the least writing effort was to publish some of the total tables from the ABC’s election computer. These category totals are not available on the ABC website but I did present them as graphics at various times on Saturday night.

But What is the “Swing” Column

The tables include a “Swing” column that needs explaining. I set up the referendum using the ABC’s computer system which is designed to analyse elections with candidates. To do this I set up two candidates in every seat called Yes and No, competing for two parties called Yes and No.

This resulted in fields on displays labelled “First Preferences”, “Two-Candidate Preferred” and “Swing” which had no real meaning. But it meant the ABC computer could output the same graphics files as at general elections, allowing the production of Yes/No television graphics and on-line election results with minimal change to software.

To try and stabilise early numbers, every polling place and electorate was set up with historic two-party preferred results from the 2022 Federal election. It meant a “swing”, really a gap, was being calculated comparing Yes% to Labor 2PP%. It is a “swing” that was never going to be uniform, but if aggregated to regions, could be used to produce a matched polling place swing that was relatively stable. This prediction was not broadcast, was not used to make automated predictions, but was useful in interpreting early figures.

The ABC coverage and on-line site at all times published simple vote tallies and percentages as supplied by the AEC. The matched swing and predicted results inside the computer were there to confirm trends and offer hints for analysis.

None of the “swings” shown below can be used as a predictive tool for the next election. This is the same as results for the 1999 Republic referendum being no guide to the 2001 Federal election.

But they provide a picture of where the Yes case failed electorally. Before the referendum my working assumption was that the Yes side could not win unless the Yes% votes could match Labor two-party preferred results by electorate from the 2022 election.

As the tables I’m about to present show, falling short of the 2022 result was exactly what happened. Setting up the history and calculating “swing” helped to confirm an early picture of the results.

Discussing this “swing”, which is really a gap, is better at explaining the results than the simple binary analysis of whether a seat voted Yes or No. As I explained in a previous post on the 1999 Republic referendum, it was the gap that explains what happened at the 1999 referendum, as it does the 2023 vote.

Results by Party

The first totals show nation-wide results broken down by the 2023 election winning party for each seat. As the column for Labor seats shows, the biggest “swing” from Labor to No occurred in Labor seats, in particular outer suburban seats in all state capitals. The swing was larger in National seats, smaller in Liberal seats, smaller again in Greens seats. There was virtually no “swing” for Others, which includes the Teal seats and the other non-Green crossbenchers. (Note I classified Queensland LNP seats by whether the member joined the Liberal or National Party in Canberra.)

NSW Regions

In broke NSW into three categories with a separate grouping for Labor’s heartland in the Hunter and Illawarra. The gap between Labor’s 2022 two-party result and the Yes Referendum result was massive in the Hunter-Illawarra, a swing that was matched in some of Labor’s safe outer Sydney safe seats.

Sydney Regions

Based on Republic Referendum results, I split Sydney up in two different ways. One was a traditional Inner/Outer divide, the second more related to distance from the coast, East versus West, a key feature of the Republic referendum. And so it proved to be for The Voice with the East-West divide showing a stark difference in the “swing” column.

Sydney By Party

The next was a simple classification of Sydney seats by party. The swing was largest in Labor seats that were mostly in the west, and smaller in Liberal seats, mainly in the east. It wasn’t the party status that explained a lot about the “swing”, it was where the seat was.

“Others” seats includes the four teal seats plus Dai Le’s seat of Fowler in Sydney’s south-west. The four ex-Liberal teal and independent seats recorded significant “swings” to Yes, while the former Labor seat of Fowler saw a big “swing” to No.

Melbourne by Region

As in Sydney, I broke Melbourne in two ways, Inner versus Outer, and East versus West, dividing seats by which side of the Yarra they lay. Inner-Outer includes both Labor and Liberal seats in each region, while East-West divides Melbourne into all Labor and Green to the west, and Labor and Liberal to the east. Both divisions produced similar differences in “swing”

Melbourne by Party

As in Sydney, Labor seats had the largest “swing” from Labor to No, the Liberal seats a smaller swing. The “Others” seats, two Teals and a Green, recorded “swings” towards Yes.

Other notable results

Of the 17 seats around the country where the Greens polled more than 20% of the vote last year, 14 recorded Yes majorities. The three that didn’t were Adelaide, Moreton and the NSW North Coast seat of Richmond.

It appears the Sydney north shore seat of Bradfield, held by frontbencher Paul Fletcher, is the only Liberal seat to vote Yes. The ‘teal’ seats had underlying Liberal two-party majorities versus Labor and five have voted Yes, with Mackellar and Curtin remaining too close to call.

Fourteen of the 22 equivalent Labor seats that voted Yes to the Republic in 1999 voted Yes to The Voice. (Equivalent as there have been boundary changes.) Eight of the 17 1999 Liberal seats that voted Yes to the Republic voted Yes to the Voice and Curtin is too close to call. Of those nine seats, only Bradfield is still a Liberal held seat with intervening losses to Labor and Teals.

5 thoughts on “A few Regional Tables on The Voice Referendum Results”

  1. Hi Antony, thank you for this fascinating analysis and all your referendum coverage. Just one question: do you know roughly what votes are still outstanding? Is it just postals and absents or are there still some pre-polls yet to be finalised?

    COMMENT: All on the day votes have been counted along with all pre-poll ordinary votes. Still to come are Pre-poll declaration votes, that is pre-polls outside of district, as well as Absent votes, provisional votes and further postal votes.

  2. Antony,

    Adding in the same sex marriage plebiscite (where I recall a bunch of Labour electorates with a no vote), is there at least some evidence for the proposition that on ‘social/cultural’ issues Labor electorates tend to be out of sync with the broad positions of the Labor Party itself to a greater degree than the seats of other parties?

    COMMENT: The same-sex marriage vote was such a result election conducted over a period of time as a postal ballot. The no vote, particularly in western Sydney, was so aligned with non-English speaking populations that I think it expresses a different set of explanations. The No side being associated with religious groups also coloured why people didn’t vote no.

  3. Hey Antony,

    While I am aware correlation is not causation, we have now seen two referendums in a row have similar generic ideas surrounding, higher yes being linked to higher education and earnings, maybe, if we see such a trend as likely or even just plausible in future referendums we can have another computation on standby, matching polling places against this the last referendum like this. Maybe this could give better and more reliable figures early on?

    COMMENT: It might or it might not. It depends on the question.

    1. Thanks for the comment. Just as a bonus, I sent you a working version of this proposal that could help in the (admittedly unlikely) event of another referendum. Thanks for the coverage of the referendum.

  4. Hey Antony, on the night of the referendum you said something along the lines of “…we’re projecting these into the forties…” when referring to NSW and VIC I believe. Out of interest, how accurate were those projected figures when compared with the result?

    COMMENT: They both finished in the 40s but lower in the 40s than appeared early on the night of the count. I had factored in that the Yes vote would be lower with pre-poll and postal votes so was confident that the Yes vote would fall from where it stabilised with election day voting. It also stabilised around where polling place matching methods were predicting.

    But the gap between polling day results and pre-poll and postal votes results was considerably wider than the 1999 referendum and wider than I have ever seen at a general election, state or federal, so the figure declined after stabilising on the polling place results. The final result compared polling day totals was down 1.6% for the Labor 2PP vote in 2022, but more than double in being down 3.8% for the Yes %.

    At last year’s federal election, the gap between Labor 2PP on polling day and with pre-polls was-3.1% and for postal -4.6%. At the referendum the same figures were double, -8.3% and -10.6%. If you look at it in Labor 2PP compared to Yes swing terms, the swing on polling day was -10.0%, for pre-polls -15.2% and for postals -16.0%. Whether you were using simple percentages or matched percentages, the early evening figures based on polling place results was the high point of Yes for the evening.

Leave a Reply