We have a number of tools to help with this task, such as the pendulum and historical swing trends, but most of these tools don’t work that well in a multi-party environment, such as the one we’re now seeing with One Nation’s surge in support.

For this blog post, I’m going to use some limited data we’ve gained on how One Nation is polling in the state seats of Victoria to model how votes might turn into seats if One Nation’s vote is higher or lower.

To do some more sophisticated modelling of how votes might turn into seats, we would ideally get an estimate of how much of the vote each party has in each seat. It doesn’t need to be perfect, but it would give a rough shape of how votes are distributed. At a national level, two MRP polls have been published this year: one by DemosAU and one by Redbridge/Accent.

At a state level, none have been published in Victoria (or any other state). But we do have one MRP that was leaked to the media.

Redbridge/Accent conducted an MRP for Victorian Trades Hall in June 2026 which was eventually leaked to media outlets including the Australian Financial Review.

This MRP reported a primary vote of 27% for One Nation, with Labor and the Coalition tied on 26%, and the Greens on 13%. It’s basically a three-way tie between the three leading parties.

This MRP reports seat totals based on these votes of 39 Coalition seats, 35 Labor seats, 12 One Nation seats and 2 Greens seats. But there’s been a lot of disagreement about those seat totals.

The MRP asked voters to fill out a full ballot, and thus used preferences as allocated by the voter. It is more traditional in polling, particularly for MRPs, to use historic preference flows instead.

While we have less data on how preferences flow with a strong One Nation vote, there are definitely some strange figures in the Redbridge/Accent MRP preference flows. In particular, the MRP suggests that Coalition voters are favouring Labor over One Nation, while recent elections have suggested the opposite.

It may be that part of the discrepancy comes from Liberal how-to-vote cards. While those may influence voters in a real election, the MRP respondents haven’t had that advice.

It makes a big difference in a scenario where One Nation makes it to the top two in numerous seats. I have produced my own preference flow estimates based on previous elections. With the same primary votes as the Redbridge/Accent MRP, the seat totals end up very different:

The specific preference flows make a huge difference. I originally had Coalition to One Nation preferences a bit higher but tweaked it after feedback from Kevin Bonham, and that dropped the One Nation seat count from 31 to 27. There is also a lot we don’t know about how Coalition and One Nation preferences will flow in Labor-Greens contests and this makes a big difference as to the Greens seat count.

Anyway, while I have questions about the MRP’s preference flows, I do appreciate the provision of primary vote figures for every seat in the state, which then allows us to do some other analysis.

So I have taken those primary vote figures for the 88 seats, and I have then applied a model where I increase the One Nation primary vote of 27% by increments of 1% until they get to a primary vote of 37%, and also decrease their vote to 17%. For each increment of 1%, I transfer one third of a percentage point from Labor, and two thirds of a percentage point from the Coalition.

I also considered a model where there was a 1:1 swap between One Nation and Coalition votes, to model the impact of changes within the right-wing part of the electorate. I suspect the Coalition contributed more of One Nation’s vote as their vote started to climb, but this model only tries to model a One Nation vote starting at 17%, at which point that early climb has finished.

So, what did I find? This next chart shows how many seats each party would win for a given One Nation vote between 17% and 37%, using my preference flow estimates and using Redbridge/Accent’s primary vote estimates as a starting point with One Nation on 27%.

I have assumed that the top two candidates make the 2CP. If two candidates are tied for second place, I have run two different counts for each of the possible 2CPs, with each result worth half a seat. This explains why some results involve a party winning half a seat.

The number of One Nation seats is very low until they get to about 22%, and then the seat count climbs relatively consistently until they get to the low 30s, then it slows down again. In this model, Labor and Coalition seat wins are roughly neck-and-neck as the One Nation vote climbs.

Interestingly, a primary vote of 32% is enough for a very slim One Nation majority. In that scenario, the Coalition is on 24.3% and Labor is on 22.7%. With quite a clear right-wing majority, it’s not surprising that the dominant conservative party would be able to win a majority in a divided contest.

This is just one model, but I think it does give some insights into how One Nation might perform in seats for a given vote.

Another feature of this modelling that I found interesting was how each party’s vote share translates differently into seat shares. Each party has a different preference position – Labor benefits from the Greens in most seats. The Coalition does a bit better than One Nation if they face off. At the recent state election in South Australia, One Nation narrowly outpolled the Liberal Party, but won one less seat. I think it’s reasonable to expect a similar primary vote for the two parties would see the Coalition win more seats.

So I went into my election model and compared how many seats each party wins on the same primary vote. The range of results is wider for One Nation than the other parties, since the model assumes a greater range of change for One Nation.

For most of the comparable range, the major parties win more seats than One Nation on the same vote.

On the current polling, with all three parties in the mid-20s, the three parties would win relatively similar numbers of seats. If any of the three parties was to boost their support into the 30s, they would then have a chance at winning a majority, because there is quite a steep dropoff in support.

Next month I plan to write seat guides for the 18 seats in South Australia, Tasmania and the ACT, to add to the 102 that have already been published. Queensland’s 30 seats will need to wait for next years.

10:30 – I have updated my federal pendulum to reflect the new boundaries for South Australian, Tasmanian and ACT federal electorates. This includes links to my guides that I’ve written so far – I am planning to write guides for these newly-completed electorates in August.

I will put together the interactive map later this morning, but that’s it for now.

10:20 – This table compares the old and new margins. There is very little change – Mayo becomes 0.4% stronger for Rebekha Sharkie, while Makin and Spence each change by 0.2% and Barker changes by 0.3%. Compared to the draft redistribution, the biggest change is in Grey/O’Donoghue, where the Liberal Party is now 0.4% stronger than they were on the draft boundaries.

10:15 – This table shows the 2PP and primary vote for each of the new seats. Only Mayo, Spence, O’Donoghue and Kingston should have differed from the draft.

10:13 – The original draft redistribution shifted 3.9% of voters – the new version shifts just 1.8%.

10:00 – The Australian Electoral Commission this morning announced the final decision on the federal redistribution of South Australia’s ten electorates.

They have adopted most of the proposed changes, but in two places they will be leaving suburbs in their original seat. A bunch of suburbs swapped between Mayo and Kingston will remain in their 2025 electorate, and the same is true for Grey and Spence. The commission has also adopted the new name of O’Donoghue to replace the seat of Grey.

This is a relatively small chance so it will be possible for me to create a map and data reflecting these changes – something the AEC has not done itself. When that’s done I will share an interactive map, my estimate of seat margins, and an estimate of the pendulum now that the first wave of redistributions has been completed. The Queensland federal redistribution is only commencing now and will take a way to finish.

11:47 – I’m going to wrap up here. Over the next few days I will produce a map showing the new boundaries, which should allow me to also produce seat guides for the Tasmanian federal electorates, but I will wait for the final boundaries in South Australia before getting to work on the next wave of my federal election guide. When that map is done, I will add it here.

11:46 – This table compares pre- and post-redistribution percentages and quotas for each group. The numbers change slightly for Bass and Lyons but all the commentary I provided in February is still relevant.

One thing I noticed while producing this table this time: Kristie Johnston has 0.7 quotas in Clark, despite only contesting about half the seat. But she also has 0.4 quotas in Lyons, which is pretty impressive considering most of the seat wasn’t able to vote for her. I suspect she would stick with Clark, but a Johnston-backed candidacy in Lyons could also be interesting.

11:33 – Here are the estimates of the percentage vote for each party group per electorate.

And here are the estimates of quotas.

Next up I’ll put up a table comparing pre-redistribution and post-redistribution figures for the main groups in each seat.

11:14 – 27.6% of voters have been moved into a new seat. That number is exactly the same as on the draft boundaries.

It’s worth remembering that the scale of the federal redistributions in this term is quite minor. Only three states and one territory are redrawing the map, and in every case the number of seats stays the same (which minimises the need for change). That means as a starting point 102/150 seats are unaffected. Two more seats (Fenner and Braddon) have now been confirmed to have no changes. The South Australian draft redistribution proposed no changes to three other seats. So this would mean 107 out of 120 seats would be unchanged, before we start redrawing the Queensland map.

11:11 – It will not be that difficult to draw a map of these new boundaries – take the draft boundaries, rename Franklin, change the Launceston-area boundary back to the 2025 boundary, and then swap Break O’Day LGA – but I won’t get that done today. Sorry for that! Should be done by Friday.

11:09 – And I just realised I don’t need to calculate a 2CP margin for Clark since it was already done for the draft boundaries.

11:07 – It will take me slightly longer to estimate 2CP margins for Clark, but this table shows primary votes and 2PP for each seat.

The Labor margin in Bass on the draft boundaries and on the 2025 boundaries was 8.0%. Now it is 8.3%. The Labor margin for Lyons had grown from 11.6% on the 2025 result to 13.1% on the draft boundaries. It is now 12.7%. There are no changes to Braddon (since the previous boundaries) or to Clark or Tongerlongeter (since the draft boundaries).

10:24 – The Australian Electoral Commission has just announced the final Tasmanian federal boundaries, although there will be further consultation on a new name. The draft version of this redistribution was published in February and you can find my analysis of those boundaries here.

The AEC media release is here and there is more information here.

The draft involved very minor changes to northern Tasmania but completely redrew the southern half of Tasmania, with Clark losing Glenorchy and picking up the rural south-west of Tasmania, Franklin losing those south-western areas and spreading up the east coast of Tasmania, while Lyons pushed into Hobart by picking up the Glenorchy area.

I have only had a chance to quickly read the AEC media release, but it appears that they are maintaining the draft boundaries in southern Tasmania, but are proposing to rename Franklin as ‘Tongerlongeter’. The AEC has given the following explanation of this name:

Tongerlongeter (c. 1790-1837) is recognised by historians as one of the most significant Tasmanian Aboriginal leaders during the early period of British colonisation. As a member of the Oyster Bay nation, Tongerlongeter’s Country is within the proposed boundaries of the electorate.

The Commission has also made some changes to the proposed boundary between Bass and Lyons. Bass will now gain the Break O Day local government area on the north-east coast, but will not gain Prospect Vale or Blackstone Heights on the south-western outskirts of Launceston.

The boundaries appear to be final, but there will be a chance for further consultation on the name Tongerlongeter.

These boundaries are also expected to be applied for Tasmanian state elections, with each electorate sending seven members to the House of Assembly. But this process is not automatic – the state Parliament will need to adopt the new boundaries once the federal process is completely finished. I will also be calculating state election vote shares once I have the maps and data, but it takes a bit longer than a federal redistribution, so probably won’t be done today.

6:32 – Unfortunately the AEC has not published maps for the final versions. I am relying on the descriptions of the borders to draw them myself – I am confident I have assigned every SA1 correctly, but it’s a bit of a pain to make. It may also be beyond my capacity to accurately do so for the Tasmanian and South Australian redistributions, depending on how much they deviate from the drafts. I will add some maps when they are done, but it probably won’t be until tomorrow.

3:30 – The scale of the changes are significantly less than on the draft boundaries. 9.9% of electors were moved on the draft, and that number is now 6.6%. On the draft boundaries, the new Fenner contained 5.9% of new electors. That number is now zero, with Fenner unchanged. The number for Bean dropped slightly from 9.4% to 9.3%, while Canberra’s changes have dropped significantly from 14.6% to 10.4%.

3:28 – And here are my estimates of the margin. My estimate of the Bean margin has become slightly smaller, falling from 0.6% on the draft to 0.4% on the final version, compared to an actual margin of 0.3% in Bean. These are all pretty loose so overall I’d say the seat is about as competitive as the old boundaries.

3:25 – Apologies for the delay. Here is my estimates of 2PP and primary votes for each electorate.

11:22 – Unfortunately I’ve just realised that the AEC has not yet published the data and maps. While draft proposals are published immediately, maps and data are not published for the final announcement until it is determined and tabled in parliament. I think I will be able to calculate the changes, but it will take a bit longer.

11:08 – The Australian Electoral Commission in the last half hour has announced the final redistribution for the ACT’s three federal electorates. In the next hour I plan to produce estimates of vote shares for the new electorates, but to quickly summarise:

• Despite original plans to make changes to the Canberra-Fenner boundary in both directions, the Fenner border is now left unchanged.
• The redistribution committee had originally proposed to move Norfolk Island and part of the suburb of Duffy from Bean to Canberra, but that has not been adopted, so those areas will remain in Bean.
• Symonston, Hume and the remainder of the Jerrabombera district will be in Canberra.

The council is elected at large, with nine councillors representing the entire council area.

In New South Wales, councils can opt in to using countbacks instead of by-elections to fill vacancies, but that process can only be used for the first eighteen months of the council term. Vacancies can also be left vacant for the final eighteen months of the term. This means that by-elections must be used for the middle year of the four year term, and that’s right where we are.

By-elections aren’t a great way to fill vacancies when multiple members represent a district. All of the voters get a chance to vote again, but only one portion of the district has lost their representative. There can be absolutely no change in the partisan make-up of voting patterns yet you can get a quite different result.

In the case of councils where all councils are elected at large (which otherwise has a lot going for it) the scale of the by-election can be far in excess of the value. Every voter in Kiama will need to vote again to replace 1/9 of the council. When there was a by-election for the City of Campbelltown in 2017, a population roughly the size of a federal electorate needed to vote to replace a single councillor.

Which is why I’m an advocate for using countbacks much more extensively, not just for the first eighteen months of the term. I would use them for all vacancies except for the last six months of the term, for all councils, and candidates should be aware of this process when they nominate.

But Kiama Council’s solution to the problems of by-elections is much worse, and shows a shocking disregard for democracy and representation.

Councillor Mike Cains resigned from council in June 2026, triggering a requirement for the council to hold a by-election.

Kiama Council is in the middle of a budget crisis, with the council needing to close a significant budget deficit with cost and service reductions. I experienced this myself when visiting Kiama on New Years Eve last year to discover that their regular fireworks had been cancelled. With this in mind, and the cost of a full council by-election estimated at $200,000, the council has instead suggested that they simply opt out of filling a vacancy.

The council decided on June 30 to ask the minister to allow the council to operate with just eight councillors for the remainder of the term and thus negate the need for a council by-election. The minister has quickly dismissed this idea as not possible. It is worth emphasising that these councillors were elected in September 2024 for a four year term. Cains resigned about 44% of the way through his term.

I think this is a disturbing reflection of how a lot of local government looks at their role of democratic representation – if councillors are simply a board of directors it makes perfect sense to just leave a seat vacant. But councillors represent different political interests and vacancies should be filled.

$200,000 may seem like a lot, but I think it is a pretty important part of local government. Without democratic representation, local government might as well be fully integrated into state government. For a council like Kiama, which is relatively small and is struggling to demonstrate its ability to financially stand on its own, it would be nice for them to show that they value their connection to their voters.

It also demonstrates the flaw in having councils directly pay for their own elections. Councils struggle with a limited revenue base and an ever-increasing demand for them to use their resources, with state and federal governments in a far more powerful position financially. It is understandable that councils would skimp on proper democratic processes. It would be far superior for the base-level costs of running elections to be covered by the state government outside of council budgets as a guarantee of democratic excellence.

Thankfully Kiama hasn’t succeeded in their push to cancel local democracy, but it is a good reminder of the problems with using local by-elections when a countback would be far superior. There is legislation sitting before the NSW parliament which would change how councillor vacancies are filled. The proposed process is convoluted – instead I would urge a change to simply make countbacks universal where they can be applied.

Labor state minister Paul Papalia today announced his retirement from the Western Australian state parliament after nineteen years in office. He has represented the seat of Secret Harbour since 2025, when the seat replaced his previous seat of Warnbro.

Papalia’s seat of Secret Harbour is at the very southern edge of Perth, and would traditionally have been considered a safe Labor seat. Papalia held the seat by an 11.5% margin in 2025, although this isn’t particularly impressive considering the statewide margin was 7.1%.

This seat is a pretty good option for One Nation. The party polled 8.4% here in the Legislative Assembly and 6.1% in the Legislative Council. This was their eighth-best result in the state, which suggests One Nation would be in a position to perform very strongly here if their statewide polling is high.

It is worth noting that, while One Nation has experienced a surge in polls in Western Australia, it has been more modest than in other states. Only DemosAU has conducted state polls here since the last election, and their last two polls showed One Nation on 17-18%.

I will obviously be following this by-election, but now you can read my guide to this by-election, here.

Tally Room member Quokka asked me to look up the primary vote figures for Katter’s Australian Party, and upon examination I realised that it was more likely that the two-candidate-preferred count in that seat would end up as KAP vs LNP, rather than LNP vs ALP, and that the seat would flip to be notional KAP.

I will explain the logic below, but in short this means that Katter’s Australian Party does not lose a seat in the redistribution, with the abolished seat of Hill being replaced by the seat of Mulgrave, and the Liberal National Party only have a net gain of one seat, not two as previously reported.

Multi-party politics makes the exercise of estimating redistribution margins far more complex. The step that slowed me down the most on Monday afternoon was dealing with seats which take in parts of electorates that had different 2CP pairings (for example, ALP vs LNP in some parts and LNP vs KAP in others). Generally my approach in these cases was to to take the 2CP which applies to the largest part of the new seat, and then calculating preference flows and applying them to the primary votes in the remainder of the seat.

I should point out that redistribution estimates can only deal with the actual votes cast at the previous election (in this case 2024). So I cannot take into account the polling surge for One Nation, nor the LNP’s gain of the Katter’s Australian Party seat of Hinchinbrook at a 2025 by-election.

In the case of Mulgrave, about 74% of voters came from the previous Mulgrave boundaries, with 26% of voters coming from the abolished seat of Hills, particularly in the area around Atherton.

The LNP won Mulgrave by a 2.7% margin in 2024, with Katter’s Australian Party coming third with 16% of the primary vote. Those parts of Mulgrave had a 1.1% LNP margin. Using those preference flows and applying them to the primary votes in the new areas, the LNP end up with a 3.2% margin.

But if you look at the primary votes in the new Mulgrave, you get a more complex picture. This table shows my primary vote estimates for votes coming from the old Mulgrave and the old Hill, and how they add up to the new Mulgrave.

This does give you a sense of the imperfections in combining votes from different seats. Legalise Cannabis and Family First were only on the ballot in one of these seats, and there was a bigger range of independents in Mulgrave. Undoubtedly, the presence of Shane Knuth in Hill would have had an impact on the size of the KAP vote in one seat over the other.

If you look at those final primary votes, it seems likelier that the final 2CP would have ended up as LNP vs KAP, not LNP vs Labor. This is the only seat where a party made the top two on primary votes but I hadn’t included them in my estimated 2CP.

When I used the same method of applying preference flows, but used the Hill 2CP flows, not the Mulgrave 2CP flows, it produced a result of KAP 6.8% vs LNP.

So I will be using this margin in my election guide, which I have started work on and should be done in about a month. That election guide will include my estimate of primary votes for each party in every seat.

It also changes the context of the overall seat count. At the 2024 election, the LNP won 52 seats to Labor’s 36, KAP’s three, and one each for the Greens and an independent. My original estimate saw the LNP gain two seats, with Gaven flipping from Labor to LNP, and the creation of Caboolture, while KAP lost the abolished seat of Hill. Labor had a net loss of one seat, with Gaven flipping to LNP, Toohey abolished and Greenbank created.

But with the reassignment of Mulgrave from LNP to KAP, it leaves KAP stable with three seats (ignoring their loss of Hinchinbrook at a by-election) and reduces the LNP net gains to just one seat.

Admittedly Mulgrave is less safe for KAP than the old Hill. Knuth won in 2024 with a 13.7% margin, and this seat has a 6.8% margin. But he can probably build on his support in the parts of the seat moving over from Mulgrave. It’s also plausible that the One Nation surge will make all of this irrelevant for KAP.

That’s where I will leave this now. I’ve amended Monday’s live blog to show my corrected Mulgrave margin, and I will be using that margin in my soon-to-be-published election guide.

These changes were relatively minor compared to the draft in March 2026. While the map overall makes things a little bit better for the LNP, that effect has not changed in today’s announcement.

Now that the boundaries have been decided, I can start work on my 2028 Queensland election guide, so I’ll be starting on that later this week.

6:40 – Here are my estimates of primary votes.

6:36 – Most margins are very similar to the draft boundaries. Labor’s margin has increased by 1.7% in Woodridge, and dropped by 2.6% in Waterford, but both seats remain very safe on paper.

The LNP needs to lose six seats to lose their majority (compared to the 2024 election). On the old boundaries, a swing of 1.9% would see the LNP lose their majority.

The LNP has picked up two extra seats on both the draft and final boundaries. On the draft boundaries, a 2.9% swing would see the LNP lose enough seats to lose their majority. On the final boundaries, a 2.8% swing would be enough.

A uniform swing of 3.5% is needed for Labor to gain a majority under either set of boundaries (or the 2024 boundaries).

The KAP seat of Hill is abolished, as well as the Labor seat of Toohey. They are replaced by the Labor seat of Greenbank and the LNP seat of Caboolture. In addition, the Labor seat of Gaven was redrawn into an LNP seat. So this produces a net gain of two seats for the LNP, a net loss of one for Labor, and a net loss of one for Katter’s Australian Party.

6:23 – Here are my estimates of margins.

Margin estimates

4:58 – Sorry I’m still working through calculating the 2CP in complex seats where the 2CP in different parts of the seat in 2024 didn’t match. Will need to wait until a bit later this evening.

3:19 – 1.8% of electors have been moved between the draft boundaries published in March and the final boundaries published today.

24 of the new seats have gained voters from another seat. Most of them are relatively small changes.

The biggest changes are on the southern edge of Brisbane. Waterford and Woodridge have each gained about 15-20,000 voters each (mostly from each other), while Ipswich West, Sandgate and Logan have all gained over 3,000 voters.

3:09 – The Commission has listed 17 seats as having a new name, with two other seats abolished and two created. However there are two examples where I think they have identified the wrong seat as being linked. There were three examples on the draft boundaries, but only two on the final version.

• Greenbank is the new seat, not Springfield. The old seat of Jordan has transferred much more substantially to Springfield than Greenbank.
• Toohey was abolished, not Stretton. The new seat of Eight Mile Plains overlaps substantially more with Stretton.

There was an issue with Waterford and Woodridge, but that seems to have been resolved.

If you take the ECQ’s word for which seats are linked, then 14.9% of electors have been moved. This number drops to 14.3% with my corrections. On the draft boundaries, this number was 14.4%. So the number of electors being moved has dropped just very slightly.

2:51 – This map allows you to compare the final boundaries to the old boundaries, but you can also toggle to show the draft boundaries. The changes between the draft and final boundaries look relatively minor at a first glance.

2:35 – I have now found the GeoJSON files, so I will probably go quiet while I put together a digital map and calculate a bunch of figures, but before I do it’s worth mentioning that the commission’s boundaries leave five seats with projected enrolment (as of 2032) more than 10% above the statewide average. These five seats are Caloundra, Coomera, Ipswich West, Morayfield and Redbank. Five seats were also above 10% on the draft, and four of them are the same. The Commission explains this as being due to those areas having very high enrolment growth, making it impossible to draw boundaries that meet both the current and projected quota.

2:20 – You can see maps on this ECQ page. Unfortunately they haven’t published digital boundary files which I’ll need before I can do my calculations.

2:15 – On further examination, this is not correct. Two seats have been restored back to their original names (Coomera and Waterford) and 37 seats have undergone boundary changes.

2:10 – The report has been published, here. It appears on first glance that the Commission has made no further changes beyond the changes in the proposed redistribution.

1:30 – The Electoral Commission of Queensland has reported that it will be publishing the final version of the Queensland state redistribution at some point this afternoon. I will post on here as soon as we have news, and will look to estimate the margins as soon as possible.

In the meantime you can read my analysis of the draft boundaries from March.

Of particular interest is the breakdown of how preferences flowed from each candidate’s voters between the final two candidates: One Nation’s David Farley and independent Michelle Milthorpe.

The AEC has also published the full distribution of preferences, which I haven’t previously written about. This gives us the three-candidate-preferred vote, and thus the flows of preferences from 3CP to 2CP (another method of calculating preference flows).

The AEC doesn’t just publish these figures for the seat overall, but also publishes them at the level of each polling place.

First up, this chart shows how primary votes for each candidate (apart from the top two) split on the two-candidate-preferred count.

It’s worth noting that the Liberal Party polled 12.4% and the Nationals polled just under 10%, but the vote for everyone else was very low. The Greens and Legalise Cannabis each polled about 2.3%.

Greens preferences flowed extremely strongly to Milthorpe, with almost 92% of Greens preferences flowing to Milthorpe. Other minor candidates were not very surprising. Right-wing minor parties generally gave strong preference flows to One Nation while independents and other minor parties favoured Milthorpe.

The Coalition candidates’ voters favoured One Nation, but not overwhelmingly so. Unsurprisingly, Nationals voters were more favourable to One Nation than Liberal voters – 69% as opposed to 59%.

The AEC’s distribution of preferences also provides us with the three-candidate-preferred count, which was:

• Farley (ON) – 44.4%
• Milthorpe (IND) – 34.4%
• Butkowski (LIB) – 21.3%

So the Liberal Party not only failed to make the final count, but they were about 13% away from such a result. If the Liberal Party had made the top two, they would have had a good chance of winning, but they weren’t close.

At this point, the 21.3% of the vote sitting with the Liberal candidate were distributed. Bear in mind that these aren’t solely Liberal primary votes. The Liberal primary vote made up less than 60% of the votes sitting with the Liberal candidate at this point. For what it’s worth, those remaining votes split 62% to One Nation and 38% to independent Milthorpe. This metric is useful because we can calculate it for every single-member election around Australia, whereas preference flows based on primary vote are only published by the AEC and the Queensland commission (and can be calculated independently for NSW state elections).

Coalition preferences absolutely do favour One Nation when they are in a race against an independent who has taken on the centre-left role in this electorate, but not overwhelmingly so. We saw this in South Australia too: in eleven races where the Liberal came third, their preferences flowed 66.6% to One Nation over Labor. In seven cases where One Nation came third, their preferences flowed 68.9% to Liberal over Labor. This could give Labor a slight advantage that adds up across many seats.

Finally, I wanted to do something with the fact that the AEC publishes preference flows by booth. I grouped the booths according to the categories I used in my pre-election guide, except I included pre-poll booths in each geographic area.

This confirms that Michelle Milthorpe did win Albury, but only slightly, with 51.6% of the two-candidate-preferred count. It was Milthorpe’s weak lead in Albury that allowed us to call the result on election night, since she needed a much bigger lead to outweigh large One Nation leads in other areas.

I wondered if there would be an interesting trend in how strongly Coalition preferences flowed to One Nation by area, and there is some interesting evidence. The Coalition preference flow is relatively weaker in Albury and Griffith, the biggest towns, but it’s also relatively weak in the rural booths of the north-east. Some of this would reflect the relative distribution of Liberal and Nationals votes but not entirely – the Nationals did much better in the north-east, but the preference flows were weaker there than in the north-west, which had a stronger Liberal vote.

Overall this is useful new data for understanding how preferences will flow, in a situation where we have very little information.

I’ve put together a full dataset of the 2026 South Australian election, including the list of polling places, candidates and vote totals down to the booth level for both houses. I have also added in three-candidate-preferred figures at the electorate level.

I have also now added the recent Farrer, Stafford and Nepean by-elections and the Tasmanian Legislative Council elections to their respective datasets.

Part of my data repository is public for everyone to access – this generally includes the most recent election for each jurisdiction. In the case of South Australia, I now have every election since 2002. I’ve added 3CP figures all the way back to 2002, and will probably gradually add that for other jurisdictions.

You can see what data is available here, and if you want the paywalled data, you can access it by signing up for $8 or more per month (+GST) via Patreon.