The Victorian state election on 28 Nov 2026 is of great interest because it is the first, state or federal, in which One Nation is a material force.
I’ve used the seat by seat betting odds from Sportsbet and TAB to simulate the outcome in the Legislative Assembly (lower house). The method is described in my post on the 2013 Federal election. The simulation was run on Nov 10. I’ll run an update just before the election.
Across a reasonable range of seat to seat correlation parameters, 35 ± 3% of paths result in a Coalition (LNP) majority. The rest have an LNP minority government with One Nation (ONP) guaranteeing confidence and supply in return for significant policy influence. The chances of Labor (ALP) forming government, even a minority one with Greens support are almost zero.
The Victorian lower house 88 electorates, so 45 are needed to form government. The current state of the lower house is 54 ALP, 29 LNP, 4 on the cross benches + 1 vacant. The cross bench comprises 2 Greens and 2 ex ALP members sitting as “independents” due to allegations of misconduct. The vacant seat is Brunswick, which was held by the Greens’ Tim Read, who died on Sep 19, 2026, so it is also a cross bench seat.
The main point is that all 5 cross bench seats are hostile to the LNP, so despite the government’s undeniable unpopularity, the LNP has a significant challenge: win at least 16 seats from Labor in the face of an unprecedented level of competition from One Nation.
Only 9 (all LNP) meet the safe criterion: paying less than 1.06 on both sites. This is a far lower proportion than previous elections. The fact that no electorates meet the safe criterion for the ALP shows One Nation is drawing support from each of the traditionally major parties. It also shows the level of dissatisfaction toward the current ALP government.
The confidence intervals for the ALP are not large: 80% of paths give 29 ± 5 seats. The Greens have a corresponding interval of [4, 7]. The 80% interval for combined ALP + GRN is [30, 39]. That is, combined, Labor and the Greens have less than 10% of paths with at least 40 seats. The chances of getting the 45 for government are very slim.
The reason adding the ALP + GRN seats increases the total, but not the variation in outcome is that success of these two parties is highly negatively correlated. The Greens are paying < 1.50 in 5 seats: Brunswick, Melbourne, Richmond, Northcote and Prahran (all inner ring Melbourne). They are paying from 2.50 to 4.00 in Pascoe Vale, Footscray, Preston and Albert Park. Of these 9 seats where the Greens are either the favourite or a material chance, only in Prahran would the Liberals beat Labor.
The number of LNP seats has a much larger than usual confidence interval, which increases with the correlation parameter. 80% of paths give 42 ± 10 seats. The reason is that the seats the LNP hopes to gain are a variety of contests: some with Labor, some with One Nation and some like Carrum and Frankston are genuinely three cornered contests.
The LNP and ONP outcomes are not as negatively correlated as ALP and GRN, but the correlation is still statistically significant. This reduces the confidence interval when we add together LNP + ONP seats: 80% of paths are in the range [49, 58]. There are a few semi-strong independents like Ian Cook in Mulgrave who could pick up a seat, although more than 1 is improbable.
Unlike the Greens, who almost exclusively take inner ring Melbourne seats from Labor, One Nation are competing to take seats from Labor, the Liberals and the Nationals. The most interesting thing is where.
One Nation is paying less than $3 on both Sportsbet and TAB in 19 of the 88 electorates.
There are 9 seats in which One Nation is the most probable winner, according to the probabilities implied by the combination of Sportsbet and TAB odds: Eureka, Melton, Pakenham, Ripon, Sunbury, Sydenham, Cranbourne, Yan Yean and Bendigo West.
I’ve listed them in that order because the first 6 have One Nation at > 50% probability of winning. The other 3 are more three cornered contests, but with One Nation still having an implied probability of winning of 44 - 49%. Only in Cranbourne and Bendigo West is the primary contest with Labor.
Add in Morwell, Macedon, Narree Warren North, Narree Warren South, Shepparton, St Albans, Wendourie, Greenvale, Mill Park and Werribee and you have the 19 electorates where One Nation is paying < $3. In 7 of these 10, the primary contest with Labor. In Macedon, Morwell and Shepparton, it is with the Liberals or Nationals.
Have a look at the map of where these electorates are and you’ll see a clear pattern: 7 of the top 9 are all outer ring metro Melbourne or its semi-rural fringe. Ripon and Bendigo West are the next electorates past the NW metro fringe.
Of the other 10, Wendourie is effectively Ballarat, Macedon is just to its north east, Morwell and Shepparton are regional towns 150 – 200 km from Melbourne. The rest are outer ring Melbourne metro, typically Labor heartland.
So all the One Nation electorates are in a ring around Melbourne. Why this geographic pattern? How much is it linked to demographics?
In answering the latter question, we need to be careful to factor out our own biases.
Analyst Anthony Green has an interesting post on the historical dynamics of One Nation preferences. Since they have been around, One Nation preferences have increasingly gone to the LNP. A reasonable conclusion is they are increasingly attracting LNP voters, although we have to be careful not to infer they are attracting people who almost always or even usually vote LNP. One Nation has historically done well in elections where unpopular Labor governments have been voted out, so they may be mostly grabbing swing voters.
The ABC has provided summary information on each of the 88 seats. Combining geographic and demographic data, the seats where One Nation are popular have properties in the submedian value range, many in the lower quartile for Melbourne. But most are mortgaged with higher than average gearing, by owners with middle to lower than average incomes. These are the people most hurt by both inflation and the RBA’s interest rate rises.
Outer ring Melbourne areas have seen demographic changes resulting from immigration, which has also been a statistically significant factor (though far from the only one) in rental shortages and property price increases. A lot of the voters there see immigration as a major cause of the problems.
The regional electorates in the semi-rural Melbourne fringe have many of the same economic issues. Although their demographics are different from outer ring Melbourne, they can see the relentless suburban growth pushing toward them and many don’t like it.
One Nation’s message on immigration is a thin veneer over: “Less Asians and definitely less Muslims and Africans”. That’s what a lot of people want. You may not, but you’re probably not voting in any of the 19 electorates where One Nation are paying less than $3.00.
One Nation is the only party whose messaging is consistently: “We hear you. You’re not wrong to say this. You should be angry. We feel the same way. We’re sick of being denigrated, lectured, or just ignored by elites. Let’s send them a message.”
Other than not letting people in who reject Western cultural values, they don’t really have any credible policies because they are a party of complaint, not of government. They are doing so well because so many people have what they feel are a lot of legitimate complaints.
In such a political environment, all you need to win a lot of votes is the above message. One Nation are the proof.
So if the major parties are historically unpopular, why isn’t the left wing party of complaint, the Greens taking more votes from Labor instead of them going to One Nation?
They are, but only significantly in the inner cities.
Most traditional Labor voters in outer suburban and regional areas are socially conservative. The dominance of antithetical social agenda in the Greens’ messaging and policies alienates them. Add to that the Greens’ predilection for insufferable, hypocritical moral lecturing and in general telling people what to do and think and they will never capture more than a small minority demographic.
A note on modelling correlation in the simulation:
For each seat k, the simulation generates a random number rk = c uo + (1 – c) uk
The u are all uniformly distributed in [0, 1]. Thus, each r is in [0, 1]. Its value determines which party wins the seat in the sample path.
uo is a single random number, common to all seats. It represents the overall electoral mood.
The uk are individual seat based electoral moods, independent from each other and u0.
Thus, c acts as a correlation parameter. Higher values cause the overall mood to swing more one way or the other, having a greater influence on the result.
Our simulation used a range of values for c = [0.2, 0.5].
The actual correlation of the rk = c2 / (c2 + (1 – c)2), which only equals c at 0.5.
One of the main results was the probability of a minority LNP government relying on One Nation support. It was p = 35 ± 3%. The probability increases with c and the greatest sensitivity is at lower values of c. For c = 0.2, p = 32%. For c = 0.33, p = 37%. For c = 0.5, p = 38%.
Values c < 0.2 aren’t realistic. Values c > 0.5 are also likely unrealistic, but given the data above, have little marginal effect on the result. So I’m confident the range for c is sufficient.
A natural question is: If One Nation are much more a factor in some areas than others, why not model local correlations instead of a statewide one?
That is, for each seat k, generate a random number rk = cS uS + cR uR + (1 – cS – cR) uk
where uS is a single, state wide electoral mood and uR are regional electoral moods, say one for inner and middle ring Melbourne, another for outer ring Melbourne and a third for regional Victoria.
We could, but the choice of cS & cR are somewhat subjective and we need cS + cR = c in our single, statewide parameter simulation.
Since the value of p above increases with c, choosing a sufficiently wide range of c values will give us essentially the same confidence interval for p as we would find modelling varying, local correlations.
A note on the 2 Labor “independents”:
The “Independent” is Will Fowles, elected in 2022 as ALP member for Ringwood. He resigned from the Parliamentary Labor Party / Caucus over sexual assault allegations, which Vic Police decided not to pursue. He was investigated by ACT Police for kicking in a hotel room door in Canberra. He claimed he was “suffering from addiction and mental health issues”. OK, but how is he fit to be in parliament?
The “Independent Labor” MLA is Darren Cheesman, member for South Barwon, near Geelong. He is a union organiser (CPSU) and ALP member, former federal ALP member for Corangamite in the Rudd-Gillard government (2007 – 2013). He was dumped from the Parliamentary Labor Party / Caucus by then Premier Jacinta Allan due to allegations of “inappropriate behaviour toward female staffers”.
No comments:
Post a Comment