FED25 – A Plethora of Estimated Margins and Problems with the AEC Version

In a pre-Christmas treat for election nerds, the Australian Electoral Commission is about to publish its estimated seat margins for next year’s Federal election.

The 2025 Federal election will be fought on new electoral boundaries following redistributions in New South Wales, Victoria and Western Australia. There has also been a redistribution in the Northern Territory though those boundaries are yet to be formally gazetted into place.

Earlier this month I published my estimates of post-redistribution margins. These are the margins I recommend the ABC should use at the 2025 election. (The post includes links to an alphabetic list of electorates and all new margins plus a new pendulum.

Having been asked my opinion of the AEC’s estimated new margins, I agree with the AEC in 137 of the 150 seats, am in close agreement on another four, but disagree on nine seats to varying degrees.

These nine seats are contests that did not finish as traditional two-party preferred races between Labor and the Coalition in 2022, and where there have been significant boundary changes.

Inside this post I have published a table that includes my estimated margins for these nine seats, as well as the AEC’s, and two other estimates published by prominent election observers, Ben Raue of the Tallyroom website, and William Bowe of the Pollbludger website.

(The AEC’s estimates are now available. You can find them at this link.)

In some entries in the table, the AEC’s estimates differ from the values that Ben, William and I calculated independently. In other entries the different methodologies adopted mean there are a range of possible estimated margins.

In this post I’ll run through detail on where there is disagreement.

But I need to make clear I am in no way suggesting there is a problem with how the AEC will conduct the 2025 election.

This Post is Not Meant as Criticism of the AEC

In writing this post, I am in no way suggesting there is anything wrong with the way the AEC will conduct the 2025 election.

My different estimates for margins have absolutely nothing to do with how the AEC take nominations, conduct postal, early and polling day voting, or count and report votes. It also does not in anyway suggest there is a problem with how the AEC reports counts on election night in its Virtual Tallyroom.

The problem concerns the more sophisticated predictive tools that the AEC provides on election night in the Virtual Tallyroom. This flows through to the feed of results provided to the media, parties and other interested users on election night.

The AEC, along with the media, parties and candidates, analyse election results based on swing, not vote tallies. We measure change from the result of the last election, not static analysis of the raw numbers.

But measuring change means you must start the analysis with a margin and a count after preferences from the previous election, including in seats affected by redistributions.

And calculating a margin requires making assumptions on how results from different electorates merge together to create margins for re-drawn electorates.

The table below sets out the different estimated margins for the nine seats in question. I have grouped the electorates by the three calculation problems that the AEC has run into in estimating new margins.

FED25 – Estimated Margins Compared

		Estimated New Margin Percentage
Division (State)	Old Margin %	Antony Green	AEC	Ben Raue	William Bowe
LIB v IND contests – can you combine Independent support for merged electorates?
Bradfield (NSW)	4.2	2.5	3.4	2.5	2.5
Mackellar (NSW)	2.5	3.3	1.8	3.3	3.2
Warringah (NSW)	11.0	9.4	10.5	9.4	9.6
LIB v IND contests – how do you account for possible Independent support in areas added to an electorate?
Wentworth (NSW)	4.2	6.8	0.6	9.0	5.8
Goldstein (VIC)	2.9	3.3	1.8	3.9	2.6
Kooyong (VIC)	2.9	2.2	2.5	3.5	0.1
ALP v GRN contests – problems in calculating the merger
Grayndler (NSW)	17.1	17.3	14.0	16.5	17.1
Cooper (VIC)	8.7	7.8	8.9	7.8	7.8
Wills (VIC)	8.6	4.6	9.0	4.6	4.6

Can can you combine Independent support for merged electorates?

At the 2022 Federal election, there were four neighbouring seats in Sydney’s north that finished as Liberal-Independent Contests.

• Bradfield – Liberal Paul Fletcher defeated Independent Nicolette Boele
• Mackellar – Independent Sophie Scamps defeated Liberal Jason Falinski
• North Sydney – Independent Kylea Tink defeated Liberal Trent Zimmerman
• Warringah – Independent Zalli Steggall was re-elected ahead of Liberal Katherine Deves.

The redistribution has abolished North Sydney with around 29,000 voters transferred to Bradfield and 32,000 to Warringah. 12,600 Warringah voters have been transferred to Mackellar.

In the above table, Ben Raue, William Bowe and I have come up roughly the same values for Bradfield, Mackellar and Warringah by adding the Liberal-Independent two-candidate preferred votes together for the new boundaries. We have made an assumption that the vote for each of these independents is interchangebale.

The AEC has decided it cannot make this assumption. In Bradfield, the AEC has decided it cannot add the vote for Nicolette Boele and Kylea Tink together, and so on with the other seats.

I’m not sure what calculation it has used, but it hasn’t combined the two-candidate preferred votes for Independents.

Where the AEC suggests that the margin of Independent Sophie Scamps in Mackellar declines, I say it increases with the inclusion of areas from Warringah.

Where the AEC says the margin of Zali Steggall in Warringah declines slightly, my calculations point to it declining further by adding vote for Kylea Tink in transferred areas to Zali Steggall.

And by including parts of North Sydney in Bradfield, I say the Liberal margin declines further than the AEC’s estimate.

All this is a disagreement about calculating a margin for the 2022 election, not a comment on how results will unfold at the 2025 election.

The problem shows up more clearly in seats that have expanded to include areas that sitting Independents did not contest in 2022.

How do you account for possible Independent support in areas added to an electorate?

In the results table, I have calculated new margins by trying to assess possible first preference votes for Independents in areas added to their seats. I base my estimates on a comparison of of House and Senate votes. Ben Raue uses an estimate based on the difference between two-party and two-candidate preferred results. William Bowe has not tried to adjust primary votes but rather allocates zero votes to the Independent and applies preference flows on accumulated primaries.

I’m not sure how the AEC calculates its estimated margins but in two of the cases it produces numbers I think make no sense.

This is clearest in Wentworth. The boundaries have been expanded to include very strong Labor voting territory from Kingsford Smith and Sydney. In my view, and in the view of Ben Raue and William Bowe, this increases the margin of Independent Allegra Spender.

The AEC estimates that Spender’s margin declines from 4.2% to 0.6%. This makes no sense whatsover.

In Wentworth at the 2022 election, the primary votes were Spender 35.8%, Liberal 40.5%, Labor 10.9% and Greens 8.3%. In the areas added to Wentworth for the 2025 election, the first preferences are Liberal 24.3%, Labor 48.4% and Greens 22.0%.

My estimate is that the Independent margin will rise from 4.2% to 6.8% based on Spender attracting much of the higher Labor and Green vote in the areas added to Wentworth, and the Liberal primary vote being lower. It makes no sense to produce a lower margin for Spender when you look at the primary votes added to Wentworth.

A similar issue comes up with Goldstein. The areas added to Goldstein from Isaacs and Hotham have a higher Labor vote and lower Liberal vote than in Goldstein at the 2022 election. Given the large decline in Labor and Greens support in Goldstein in 2022 with the presence of an Independent, I think it is safe to assume the same in relation to the areas added to Goldstein for the 2025 election.

Kooyong is much more complex. 26,000 voters have been added from the abolished seat of Higgins. This area had a relatively high Liberal primary vote and I estimate it produces a slightly decline in the estimated Independent margin for the new Kooyong. It is a tough one to estimate.

In these three cases the AEC made a decision to make no estimate of how the Independent would go in the areas added to each seat. But by not making an estimate, and instead just using an incomplete estimate of votes, they produce margins that will be mis-reported.

The AEC’s margins will cause news stories to report that the margin of Independent Allegra Spender in Wentworth has been slashed and the Zoe Daniels’ hold on Goldstein has been weakened. In reality, the position of the Independent will be improved in both seats.

The AEC has decided to make no assumption about how voting patterns will change in the areas added to Wentworth and Goldstein. But by being neutral in not making assumptions, but still producing margins, they will create a false impression that margins have been sliced.

Again, this is not a criticism of the AEC and has nothing to do with the conduct of the next election. But it is an interesting example of how making neutral assumptions can create false impressions by missing key points about the data being analysed.

Problems with Labor-Green Contests

When it comes to Grayndler, Cooper and Wills, there seems to be something wrong with the AEC’s calculations.

The areas gained by Grayndler from Watson and Barton have a higher Labor vote, higher Liberal vote, and lower Green vote, than the areas of Grayndler continuing on from the old boundaries. The Labor margin in Grayndler must increase. It cannot decrease as suggested by the AEC.

In Cooper, the whole of the old Cooper is retained on the new boundaries, and Labor polled 58.7% in the seat in 2022. Areas have been added from Melbourne where Labor polled 39.5% after preferences versus the Greens in 2022. The Labor margin must decline as I calculated. It cannot increase as suggested by the AEC’s estimate.

This is even clearer in Wills. Labor polled 58.6% in the seat in 2022. Around 7,300 voters in an area where Labor’s vote was 64.8% has been removed, while into the seat from Melbourne come around 16,500 voters in an area where Labor polled only 37.2% after preferences. Labor’s margin must decline and my estimate is Labor’s two-candidate preferred result would be 54.6%. Improbably the AEC suggest it rises to 59.0% which makes no sense whatsoever.

A Final Point

People are free to assume whatever margin they like as a starting point for the next election. After all, adopting a margin is just theorising on the impact of new boundaries on 2022 election results.

The AEC has found itself in the difficult position of being neutral and making no assumptions about voting patterns at the last election, but as a result producing margins that make little sense in several seats if you look at the change in first preference support in areas combined to be part of new electorates.

In the end this is a disagreement about the past, not about the next election.

A discussion about margins should in no way raise concerns about how the AEC will conduct the 2025 election. I have no doubt it will conduct the 2025 with all the neutrality and diligence it has done in the past.

In this post I hope I’ve explained the process and why using the AEC’s estimated margins alone could create a mis-leading impression of the margins in the nine seats I’ve mentioned.

People can use to choose the AEC’s margins in these seats, or my margins, or their own margins if they wish. The starting margin affects the swing you get on election night and any prediction based on swing. But the starting margin has no impact on the actual result of the next election which is decided by the tally of votes in 2025, not calculations of adjusted margins for 2022.

In these nine seats in particular, no one should treat the AEC’s margins as something chiselled in stone and brought down from a mountain.

There are assumptions behind the AEC’s estimated margins, as there are behind my estimates and those of others.

In this post I hope I’ve explained those assumptions before everyone adopts the AEC’s margins as gospel.