US Midterms 2026

The US Midterm elections are upon us, and advocates for both major parties are framing this election as one which will enable or inhibit the Trump agenda for his remaining two years in the presidency.

House Rule

Analysis across the board expects the Democrats to re-take the house. My analysis unsurprisingly agrees. There are 435 voting members of the house, and the Republicans hold the balance here by a razor-thin margin of 215:210. If Democrats can flip three seats from their 2024 result, they’ll hold the majority. Several factors playing against the Democrats in 2024 (grievances against the Biden government, rapid installation of Harris without a primary or sufficient campaign time etc.) are no longer in play; conversely grievances against the Republicans have exploded as a result of the war in Iran, inflation and gas prices, perceptions of corruption and authoritarian-style government overreach among myriad other issues. At this stage if the Republicans only lost three seats, it would be a cause for celebration among the party.

In the analysis below, I will also include the non-voting delegates in the house representing American Samoa, D.C., Guam, the Northern Mariana Islands, the US Virgin Islands and Puerto Rico – although Puerto Rico’s Democrat representative is not actually up for election in 2026, so that’s already one seat of 441 secured on the board for the Dems.

Some other races are also foregone conclusions where a representative is running unopposed or one of the major parties is not contesting the seat at all. This lines up a further 18 seats for the Ds (Arizona 3; California 4, 7, 11, 12, 14, 29, 34 and 37; Florida 10; Massachusetts 1, 2, 5 and 7; New Jersey 8; Pennsylvania 3; Wisconsin 2; and the non-voting seat of US Virgina Islands) and two for the Rs (California 40 and non-voting Northern Mariana Islands).

Given the heavy fall in Republican support, I am comfortable also assuming no blue seats are turning red, except where these have been caught up as part of the controversial mid-decade redistricting of seats as part of a blatant Gerrymandering campaign by both parties.

Finally, in this first-pass prediction, I am assuming certain die-hard Republican seats are unwinnable for Democrats regardless of the political climate, but that Dems will win all typically Democrat seats. Using the Cook Partisan Voting Index, which measures each district’s deviation from the national voting average, I am assuming any seat with a Republican lean of R+15 or greater will vote red, while any Democrat lean (D+1 or higher) will vote blue. The threshold of R+15 was determined by (A) taking the the polled national vote split 43:52 which reveals a 9-point gap, (B) multiplying this by 150% for additional safety (rounded to 14) and (C) adding one. In other words, because the national vote has a 9-point lean to the Democrats, seats 9 points to the right of the national average are considered boarder-line and in-play, and a district needs a lean of over one-and-a-half times this to still be considered safe for Republicans.

Combining the foregone conclusions, re-election of all Democrats and the awarding of safe Republican districts, the first prediction looks as follows:

Prediction 1: Republican 74+1, Democrat 209+3, Undecided 152+2

One immediate peculiarity here is Florida – a strongly Republican state which has only one ‘R’ so far, and four ‘D’s. I alluded to the rash of out-of-season redistricting attempts ahead of this election; Florida was one of the states where this occurred. Because winning a district only requires one more than 50% of the vote, and winning 51% is as good as winning 99%, the goal in Gerrymandering is to spread your own supporters out to secure small margins of victory in many seats, while consolidating your opponents in a few incredibly strong seats. This almost certainly accounts for the many Florida seats falling under the R+15 threshold.

In the states where courts have allowed redistricting since 2024, I’ve assumed that the Gerrymandering in Republican controlled states was conducted competently to preserve all Republican incumbents (and new red seats), even if these maps were drawn prior to a substantial downtick in Republican support.

Pro-Democrat Gerrymandering has also occurred (e.g. in California) but because any seat with a nominal D+1 Cook Partisan Voting Index has already been given to the Dems in our prediction, the only seats that we are adding on at this stage are Republican – 64 in all (Alabama 2; Florida 2-8, 11-13, 15-19, 21, 22 and 26-28; North Carolina 3, 5-11, 13 and 14; Ohio 5, 7, 8 and 10; Tennessee 4-8; Texas 2-6, 8-10, 12, 14, 15, 17, 21-27, 31-32, 35, 36 and 38).

Prediction 2: Republican 138+1, Democrat 209+3, Undecided 88+2

This gives us a "short"-list of 90 districts for more granular analysis. Polling is available for many of these, aggregated here. Where I have subjectively considered the polling to favour one candidate or another, I’ve marked the board accordingly below. (This is necessarily subjective, because it involves weighing polls not only on how many found Republican or Democrats to be leading, but also by how strong the lead is and how recent these polls are). Republican-favouring polls from before the start of the Iran war (and subsequent collapse in Republican support) are given little weight. Some seats with only one poll might still be given to one or the other party if the poll is recent enough (especially given we’re running out of meaningful other data); other districts with two or three polls all from last year may be ignored.

From my judgement of the polls, Republicans pick up 14 more seats (Alaska at-large; Colorado 4 and 8; Indiana 5; Kansas 2; Minnesota 8; New Jersey 2; New York 21 and 23; Oklahoma 5; South Carolina 1; Tennessee 9; Virginia 5 and 6) and Democrats 24 (Arizona 1 and 6; California 6; Colorado 5; Florida 14 and 25; Iowa 1 and 2; Kentucky 6; Michigan 4 and 7; Minnesota 1; Missouri 2; Montana 1; New Jersey 7; Ohio 1 and 15; Pennsylvania 7, 8 and 10; Texas 28 and 34; Virginia 1 and 2).

There was some obvious overlap between the states for which polling was available and those we had not called yet. That said, some uncalled races had no polling for the general election. Perhaps naively, I assume the fact that no-one is watching those races closely is that they are considered pretty clear cut, which would favour the Republican incumbents in those 32 seats (Arizona 5 and 8; Arkansas 3; California 5; Georgia 1, 10 and 12; Illinois 16; Indiana 2; Kansas 4; Michigan 1 and 5; Minnesota 6; Mississippi 3; Missouri 3; New Jersey 4; New York 2, 11 and 24; Ohio 14; Oklahoma 1; Oregon 2; Pennsylvania 11 and 16; South Carolina 2, 4 and 5; Washington 4 and 5; Wisconsin 5, 7 and 8). Adding these in, along with a safe bet of a Republican returned in American Samoa and a Democrat in Guam, we get an updated prediction as follows: 

Prediction 3: Republican 185+2, Democrat 234+4, Undecided 16

And that’s really where the “serious” House prediction ends – but for the fun of calling every tossup, let’s assume the remaining 16 Electorates split evenly and sort them based on their Cook Partisan Voting Index:

Conclusion

Giving a final prediction of:

• Republicans 193 + American Samoa and Northern Mariana Islands
• Democrats 242 + D.C., Guam, US Virgin Islands and the continuing term in Puerto Rico

Send in the Senate

This is where things get interesting – while virtually every credible source has the Dems easily winning back the House, the Senate is a lot closer. In 2026 we have 35 senators up for election – 33 at the expiry of their six-year term and two special elections to replace Senators promoted to Trumps executive branch (to the roles of Vice President and Secretary of State).

The remaining 65 seats are shared between 31 Republicans, 32 Democrats and 2 Independents who caucus with the Democrats and, for balance of power purposes, are treated as Dems below.

The Republicans face a nominal disadvantage on this make-up, needing 20 seats to retain power vs the Dems’ 16. The Republicans currently hold 22 of the seats up for election, meaning they can only lose two and maintain an outright majority, or lose three and rely on their Vice-Presidential tie-break.

While the Republicans have a lot to lose in these elections (control of the Senate in addition to near-certain loss of the House), the stakes are also high for the Democrats. Despite consistently close polling in many races making this neck-and-neck, there is an expectation that the Democrats can and will win both houses – failure to do so will appear as a damning indictment regardless of how well they perform in absolute terms.

First let’s look at the 2026 senate races, 27 of which have been consistently rated across twelve separate publications as at least leaning to one party or another (14 to Republican, 13 to Democrat):

To this I would comfortably add South Carolina for the Republicans, as only Fox and RCP considered this a tossup, and Michigan for Dems, which has four publications considering it a tossup but I just don’t see a seat won by Dems in 2020’s Trump-returning-victory year falling to the Reps in 2026 given… everything.

That gives us six key races (Alaska, Iowa, Kansas, Maine, Ohio and Texas) of which the Republicans need to hold four to manage a tie in the Senate.

Alaska

In Alaska, polling consistently (though marginally) favours Democrat Mary Peltola, including every ranked choice poll (which is how the election will be conducted) as well as eleven of the fifteen first past the post polls since the primaries. (The remaining four include two ties, and two Republican wins, all from Republican-favouring polling houses.)

On this basis, Alaska seems like a Democrat win.

Iowa

In Iowa the polling is evenly divided. Two polling aggregators favour Republican Ashley Hinson and two favour Democrat Josh Turek. Eighteen individual polls favour the Republicans, and 18 favour the Democrats – though two Republican leaning polls were conducted before the primaries versus only one for the Dems.

This could easily go either way, but with a higher hill to climb and a Republican currently on-seat I’m nominally turning this one red.

Kansas

Despite not having won a Kansas senate race in almost a century, the Democrats lead the polling this year.

Kansas will be close, but I’m tipping the Democrats here.

Maine

Democrats are dominating the polling in Maine…

Ohio

… and in Ohio…

Texas

… and in Texas.

Conclusion

While US politics have taught me more than once to not trust blindly in polls, that brings us to 16 Republican wins and 19 Democrats, enough to flip the Senate into Democrat control. Even if some of these go against the polls, there is still a solid path to victory for the Dems.

South Australia 2026

So here we are, on the eve of the South Australian state election. In some respects, this looks like a very straightforward race to predict; in others it's quite unusual. Polling across the board agrees that tomorrow should be a resounding success for the Labor state government. On the other hand - and by no means uniquely in South Australia - the long standing dominance of the Coalition parties appears to be heavily threatened by a large One Nation primary vote.

A Couple of Brief Excuses

It's pretty typical around these here parts for me to focus overwhelmingly on whichever house has the multi-electorate races partly because of the granularity that data often provides and mostly due to time constraints - I give myself extensive latitude to procrastinate with the excuse that you never know if more polling might precipitate from the aether.

In this election, though, there are two other "houses". One is the First Nations Voice. However, due to a lack of attention from media outlets, there is no polling I can find. Further, the vast majority of candidates do not have any party affiliation. So my excuse for not spending more time on this race is simply that I have no meaningful data to play with.

In the case of the Legislative Council, my excuse is more typical. The state-wide proportional representation (an increasingly popular approach to upper house elections) makes the predictive math relatively simple. (I say simple, not necessarily accurate.)

There are 22 seats to fill, and due to the staggered terms, 11 will remain in place:

For the remaining 11, the latest state-wide polling might be applied. The last upper-house-specific polling I saw came from 23 February and had Labor at around 33.8%. (N.B. this was reported as 38%, but excluded 11% of undecided voters; 38% or the remaining 89% is just over 33.8%). One Nation polls just under 18.7%, Liberals just under 13.4% and Greens around 9.8%.

More recent polling is, of course, available for the lower house. Even if this is representative of upper house voting intention, it doesn't shift the dial that much: Labor 34.4%, One Nation 21.4$, Liberal 15.8% and Greens 10.2% (after accounting for an excluded 7% undecided).

Any party polling over 1/12 (8.333%) of the vote is secured a seat, so this locks in 4 Labor seats, 2 One Nation, 1 Liberal, and 1 Green.

With a remaining primary vote between and 5.1% and 7.5%, the Liberals are well lined up to pick up another seat on preferences. The last two are a wild guess, with around one seats worth of support strewn across minor parties which could flow in any direction and another seat remaining in the "undecided" camp.

Nevertheless there is no mathematical possibility for the Liberal party to take a majority, and the possibility of Labor picking up the last two spots to secure 11/22 is purely theoretical. That means a hung upper house with Labor as the largest player, and a path to majority on any vote either with the Greens and one cross bencher to their left, or the Liberal party to their right.

The main game

We now turn our attention to the lower house. And in that "we" I of course include Antony Green, who published an electoral pendulum for the Legislative Assembly last year. I've altered this slightly, reverting Black and Dunstan back to pre-by-election margins, and dropping out a some against Independents including the seat of Mt Gambier which is vacant following the resignation and imprisonment of it's previous incumbent.

(I have assumed Kavel will return to a 2-party race with the retirement of Cregan, and while this year is a good year for a Liberal-turned-Independent to hold their seat, this is the first Independent race for MacKillop's McBride so it's easiest to treat the seat as Liberal for now.)

Now typically, we'd calculate the swing from last election to the latest two party preferred poll (54.6:45.4 to 59:41 is a swing of 4.4 percentage points to Labor) and apply that uniformly here.

But in this election, the Liberal Party isn't necessarily the party to be looking at in two party preferred terms. One Nation also polls 59:41 against Labor. Does this mean that We can construct a hypothetical ALP v ONP pendulum by just replacing all Liberal numbers with One Nation? Probably not. Do I have a better way of calculating such a pendulum? Nope.

But that same polling does suggest the Liberal primary votes exceed ONP in "Inner Metro Adelaide". I can't find a strict definition of how that term was used in the polling, but I assume that includes Bragg, Colton and Morphett. The other six? That'll come down to preferences.

Here's what we know on a more granular level: In Chaffey, One Nation came in third in 2022 despite the party's reasonably lacklustre showing across the state. Schubert saw a fourth place finish, behind the two major parties and the Greens who we may assume have negligible overlap in constituencies. In Finniss and Hammond, Independent candidates were more successful (though, of course, ONPs credibility as a viable third option has shot up significantly since then).

Chaffey certainly seems the most vulnerable to a ONP takeover, followed by Schubert. MacKillop feels ripe for an IND hold-out, as do Mt Gambier where we have the only district-specific poll, Narungga and Stuart. None of these are credible wins for Labor; the fall in Liberal popularity means they won't return to the blue; and the viable option of an independent largely removes the push-factor driving people to One Nation.

Final Prediction

A Very Late National Prediction

 Normally we take the polled swing from last election, add that to the pendulum and start from there. This year, though, the polling is basically exactly where the election landed - negligible swing on a two-party-preferred basis!

There's a little more variation at the primary vote level, but even that has the 2022 result well inside the recent variation:

That's boring - our prediction is just a copy-paste of the 2022 results.

So, let's try something else. At the very least, we don't seem to be looking at much of a swing, so we can lock in any seats classified as "fairly safe" (i.e. margins > 6 percentage points):

Next, we're luck to have a lot of seat-by-seat polling. Let's add in everyone where that polling has been giving a consistent result: 

And lastly, let's look at the history of the much-more-numerous-than-usual remaining seats, and a general gut feel, to finish us off (the reason this post is so down-to-the-wire):

Banks – LIB

Traditionally a Labor seat, Banks was won in 2013 by David Coleman who continues to contest the seat. Having weathered the swings back and forth since the end of the Rudd-Gillard years, the Liberals are probably resilient enough to hold on here.

Bass – LIB

While less robust for the Liberals in modern time, Bass has also has a less stalwart Labor history in its early days. While this will surely be a close one, I think Bridget Archer’s incumbency and the lack of expected major swings in general can hold this one for the Liberals.

Blair – ALP

Shayne Neumann won Blair with the Kevin Rudd landslide of 2007 and hasn’t lost it over the following almost-two-decades. Running again, that streak could well continue.

Bradfield – IND

Bradfield has been Liberal since its inception in 1949. It’s once of those seats where Labor has such a low chance of winning, it’s feasible for an independent to out-place them on primary vote, pick up their preferences and then vault into a winning position. This nearly happened in 2022 for Nicolette Boele, who is contesting the seat again this year. With, if anything, a slight swing this year to minor parties and independents, she may well take the seat this time.

Brisbane – ALP

Brisbane has moved about historically between Labor and Liberals, and last election the Greens finally won. As an inner city seat, it is a strong option for the Greens, but the major parties may have seen it as winnable too and put a lot of resources in here. If so, the Australian pattern of one-term Geens members may continue, and their preferences when defeated would likely get Labor over the line.

Bullwinkel – LIB

Bullwinkel is a new seat, but is made of parts of Hasluck and Swan, with mixed voting histories, as well as more rural Coalition strongholds of Canning, Durack and O’Connor. Though nominally held by Labour this is untested, and at the recent state election a number of seats in this area were won by Liberals or Nationals despite a strong Labor result elsewhere.

Calare – IND

Calare has been won by Nationals and Independents since 1996. In 2022 it was won by Andrew Gee for the Nationals against Independent Kate Hook. Now Gee is an Independent, Hook is running again and the Nationals have fielded Sam Farraway, making for an interesting three-way race. Preferences will be unpredictable and chaotic. Polling in the seat suggests the Nats can beat Hook again, but will struggle to beat Gee.

Casey – LIB

Liberal since 1984, and with 2025 not poised to be a year of shock upsets, expect the Liberals to take this one.

Chisholm – Tossup

History gives little instruction in Chisholm. I’ve tried not to do this in recent years given the ultimately low stakes on this blog, but I can only call this with a coin-flip so let’s call it as it is – a tossup – and hope it doesn’t come down to this seat in the final prediction.

Cowper – NAT

National since 1963. National again. Only Independent Cas Heise stands a chance at unseating them, and didn’t do it in 2022, so will have to fight very hard to pull it off this time around.

Curtin – IND

The seat has only ever been Liberal or Independent. Independent Kate Chaney took the seat last election, and with greater name recognition and a high media profile, she’ll likely keep it.

Deakin – ALP

The closest seat in the race, if any seat will flip this is the one to bet on. So I will – ALP to take Deakin from the Liberals.

Dickson – LNP

It’s a close one, it’s a hugely contentious one, but ultimately the Liberals will have put in whatever resources it takes to keep their leader, Peter Dutton, in his seat.

Forrest – LIB

Liberal since 1972, there is no real suggestion ALP can take this. Some speculation has been put behind prominent Independent Sue Chapman, but I think that’s a tough ask this year when the 2PP is still considered to be Liberal vs Labor.

Fowler – IND

Suggestions the Labor could regain the seat from Independent Dai Le are not unreasonable, but certainly the less likely outcome in my view.

Goldstein – IND

Likewise, Liberal winning seats back from Independent Zoe Daniel is a big ask.

Griffith – ALP

A three-way race between Labor, Liberal and Greens will be unpredictable, and all three parties want this seat. Ultimately, if Labor falls out first, they’ll preference Greens, while you’d expect both Greens and Liberals to preference Labor over each other, giving Labor the upper edge in a game of pure chance (which an election never is).

Hunter – ALP

Labor since 1910. Labor Again.

Kooyong – IND

I don’t think the major parties have done themselves any favours this year in trying to oust Independents like Monique Ryan.

Lingiari – ALP

A tight one to be sure, but Lingiari has been Labor, holding out against the Country Liberals, since it’s start in 2001.

Lyons – Tossup

What can I say. It’s a tough one.

Mackellar – IND

The fight between Liberal and Labor has probably reduced both parties in the eyes of voters, which will continue to play well of incumbent Independent Sophie Scamps.

McEwen – ALP

McEwen’s a weird one historically. Labor couldn’t win it in Rudd’s first popular election, but then did under Gillard. They held on in the swing back to Liberals and hold it still. Given this tenacity, I don’t expect a change this year.

Menzies – LIB

A Liberal seat named for the Liberal founder that’s been Liberal since it was created.

Monash – IND

The mixed polling here arises from the consistent result that polls framed as Labor v Liberal see the Libs win, and those framed as Liberal vs Independent incumbent (and former Liberal) Russell Broadbent see the latter squeeze ahead with a margin of a few percentage points.

Paterson – ALP

A mixed history to be sure, but Meryl Swanson has built a decent support base over her time in office.

Reid – ALP

A Labor seat with only a small deviation in the 2010s, it’s a strong chance for Labor. Certainly not a sure thing, but the only reason this wasn’t a clean sweep on the seat-wide polling was a single 50-50 split result last year.

Robertson – ALP

A bit of a bellwether seat, given the national results are expected to return Labor in some form, let’s say the same for Robertson.

Ryan – GRN

I’ve been pretty down on the chances for the Greens so far (see Brisbane and Griffith), but they really out-did Labor here last time, and they’ll be sure to scoop up the lion’s share of preferences if that repeats.

Shortland – ALP

Labor since 1949, like Reid the only polling saying otherwise was reflecting a single 50-50 split.

Sturt - LIB

Liberal since 1972. The most recent local polling suggests a change here, but I’m not convinced.

Tangney - LIB

Last election was the first time Labor won this seat since 1984. Local polling tends to be breaking their way, and the power of incumbency shouldn’t be underestimated, but if I always played by the numbers there’s be no point commenting at all, so I’m going to take the long-shot and say the Libs could regain this one.

Wannon – LIB

Liberal since 1955, it’s a tall ask to unseat them here. There is, to be sure, some very strong polling for Independent Alexander Dyson – up to 61.9% in one case – but I’m not sure that’ll materialise on the ground today.

Werriwa – ALP

Labor since 1934. Come on.

Final Prediction

That's Labor 74, Coalition 58, 12 Independents (!) plus Katter and Sharkie and two Greens. With two tossups, Labor could reach 76, but minority government is more likely. With such a large cross-bench, though, Labor has plenty of options.

Bonus: WA Electoral Maps

It's been a long while since I last did one of these, but the official maps are not very useful for those who like to fill in a cop at home.

I've made these ones for my personal use, so I figures I might as well post them here for others who may be interested. There is an anti-aliased version to print out and annotate by hand, and an aliased version for use with basic image software.

Anti-aliased:

Aliased:

New WAs of Voting

Western Australia has changed how it elects its Legislative Council, now adopting a single, state-wide electoral district which elects 37 Councillors through proportional representation. For a Councillor to be elected, then, they need a quota of 1/38^th of the votes +1 (disregarding any remainder) or ~2.63% of the vote.

There isn’t, to my knowledge, any polling on people’s voting intentions in the 2025 Legislative Council election. However, historically, voting in the Council and Assembly has been generally comparable.

Notes: This chart only plots parties running in 2025 where they ran in the Legislative Council and Legislative Assembly in the same year.

This is somewhat surprising, as I would have expected the plethora of candidates on upper-house ballots to entice more voters away from the major parties who make up the predominant options in the lower-house.

Given there is little evidence that people vary their voting preference between the Legislative Assembly and Council despite radically different electoral processes, it is fair to assume the relatively minor change in counting method will not significantly divert the average voter.

The equivalence between voting in the two houses only seems to break down significantly when polling below about 1% or 1.5% where parties can perform better in the Council elections with poorer, or virtually absent, Assembly success; this may be a symptom of having candidates in only a handful (or less) Assembly races but access to much larger voter groups in the Council, a phenomena greatly exaggerated under the new system where state-wide exposure is available to Council candidates.

It is also worth noting that polls in the leadup to elections do not tend to overestimate “Other” voters, which one might assume was the case if undecideds weren’t attracted to any major party.

Note: Other here is defined as any party other than Labor, Liberal, National or Greens.

Current polling has Labor at 42%, Liberal at 32%, Greens at 12%, One Nation at 4% and Nationals at 3% of the primary vote.

From this we can safely assume on first preferences, with a little flexibility that’ll all wash out in the preference flows: 16 Labor Councillors, one National, 12 Liberal, four Greens with a fifth well on the way, one One Nation and a good start on a second.

That alone accounts for 34 of the 37 upper house seats, and would require Labor to pick up all three remaining to secure a majority. It’s still a very workable upper-house for Labor without this, though, with the ability to secure a clean majority on left-of-center issues by working with the Greens or on core business-as-usual matters by working with the Liberal party (e.g. supporting employment opportunities and keeping the economy afloat) if the minor parties start becoming too “activist” for the government’s liking.

There are three independents currently in the Legislative Council running in 2025: Kingston, Moermond and (following a recent legal name change) Trump. I don’t see all of these holding onto their seats. A high profile in the South West likely won’t carry any of these, though Mr. Aussie Trump may secure some support from fringe voters enamoured with the state of US politics.

Similar at-the-ballot impressions may favour the Animal Justice Party, Australian Christians, Legalise Cannabis, Stop Pedophiles! and to a lesser extent Libertarian and Sustainable Australia Parties. Shooters, Fishers and Farmers may also get support in key regional areas where they’ve built up a presence.

I’d suspect one of these to cut through on luck of the preference draw (taking a wild guess, based on past successes, Legalise Cannabis), Greens to get their fifth seat with aid of Animal Justice or Sustainable Australia, and One Nation to get their second from preferences of Libertarians and Australian Christians.

WA: The (Polling) Pit and the Pendulum

Western Australia goes to the polls next weekend with what looks like an easy win for Labor. The party has consistently polled 10 points ahead on a two-party-preferred (2PP) basis since 2023 and holds 53 of the Legislative Assembly’s 59 seats.

However, some caution is required; the 2021 electoral results were obviously an aberration deriving in part from a very successful early response in WA to the COVID19 pandemic that largely kept the state open while the east coast suffered with frequent lock-downs and outbreaks. Labour not only secured about 70% of the 2PP vote in 2021, they reduced the seats held by the Liberal party to an extent that made them the minority member of the Coalition. I am not aware of any serious observer who considers this to reflect a new status quo in Western Australian politics, and it would be folly to assume that an electoral prediction could simply draw on 2021 data as a starting point.

Simply drawing on 2021 data as a starting point

We can calculate the anticipated state-wide swing to the incumbent government by subtracting the previous 2PP electoral result (69.7%) from the latest polling numbers (56%) which comes out to -13.7 percentage points.

We can then apply this uniformly to the pendulum calculated by Antony Green and get a feel for which seats are likely to flip.

DISTRICT	INCUMBENT	ALP MARGIN	-13.7 PP
Albany	ALP	11.0%	-2.7%
Armadale	ALP	33.8%	20.1%
Balcatta	ALP	25.8%	12.1%
Baldivis	ALP	35.8%	22.1%
Bassendean	ALP	31.7%	18.0%
Bateman	ALP	6.7%	-7.0%
Belmont	ALP	29.4%	15.7%
Bibra Lake	ALP	28.2%	14.5%
Bicton	ALP	16.2%	2.5%
Bunbury	ALP	22.5%	8.8%
Butler	ALP	32.6%	18.9%
Cannington	ALP	30.7%	17.0%
Carine	ALP	3.9%	-9.8%
Central Wheatbelt	NAT	-9.3%	-23.0%
Churchlands	ALP	1.6%	-12.1%
Cockburn	ALP	28.1%	14.4%
Collie-Preston	ALP	23.3%	9.6%
Cottesloe	LIB	-7.4%	-21.1%
Darling Range	ALP	14.1%	0.4%
Dawesville	ALP	13.1%	-0.6%
Forrestfield	ALP	22.9%	9.2%
Fremantle	ALP	N/A	N/A
Geraldton	ALP	9.3%	-4.4%
Girrawheen	ALP	33.4%	19.7%
Hillarys	ALP	18.7%	5.0%
Jandakot	ALP	18.2%	4.5%
Joondalup	ALP	25.5%	11.8%
Kalamunda	ALP	14.5%	0.8%
Kalgoorlie	ALP	11.2%	-2.5%
Kimberley	ALP	21.4%	7.7%
Kingsley	ALP	16.9%	3.2%
Kwinana	ALP	38.2%	24.5%
Landsdale	ALP	25.6%	11.9%
Mandurah	ALP	24.2%	10.5%
Maylands	ALP	29.3%	15.6%
Midland	ALP	26.9%	13.2%
Mid-West	NAT	-8.6%	-22.3%
Mindarie	ALP	27.9%	14.2%
Morley	ALP	27.9%	14.2%
Mount Lawley	ALP	21.9%	8.2%
Murray-Wellington	ALP	17.3%	3.6%
Nedlands	ALP	3.1%	-10.6%
Oakford	ALP	27.7%	14.0%
Perth	ALP	29.2%	15.5%
Pilbara	ALP	17.6%	3.9%
Riverton	ALP	10.9%	-2.8%
Rockingham	ALP	37.7%	24.0%
Roe	NAT	-12.2%	-25.9%
Scarborough	ALP	9.5%	-4.2%
Secret Harbour	ALP	31.3%	17.6%
South Perth	ALP	10.1%	-3.6%
Southern River	ALP	33.1%	19.4%
Swan Hills	ALP	27.3%	13.6%
Thornlie	ALP	30.6%	16.9%
Vasse	LIB	-4.3%	-18.0%
Victoria Park	ALP	27.7%	14.0%
Wanneroo	ALP	28.5%	14.8%
Warren-Blackwood	ALP	2.2%	-11.5%
West Swan	ALP	34.8%	21.1%

Notes:
Freemantle is marked N/A, as the 2PP result in that seat was a Labor v Greens race. 
Unsurprisingly most of the seats leaving the Labor fold were held by the Coalition before the 2021 election, and here are assigned to the Liberal or National parties based on the previous incumbent.
Geraldton is given to the Liberal parts, as they won the seat in the 2017 election, though the member defected to the Nationals in 2019.
Albany is given to the Liberals as this party came in second place in 2021 and held the seat in 2001; the Nationals last held Albany in 1956.

On these numbers the Coalition picks up 12 seats (11 Liberal seats and 1 National). This count is hardly surprising; the current 2PP polling (56:44) is quite close to the 2017 electoral results (55.5:44.5) forming the Legislative Assembly prior to the 2021 upset which cost exactly this distribution of seats.

That leaves the Labor party with 47 seats, ignoring Fremantle for the time being, and a safe enough margin to secure majority government even if they also lose the close fights in Bicton, Darling Range, Kalamunda and Kingsley. The Coalition only holds 17 (Liberal 13, Nationals 4).

Forewarning – this comfortable lead will not be substantially shaken by the analysis through the rest of this post.

Back to the Future

Perhaps, if the 2021 election was such an aberration, it is more appropriate to ignore that year entirely and compare current polling to the pendulum as it stood before that date.

As already noted, there is little movement between the 2017 results and the 2025 polling. A uniform swing of +0.5 percentage points is well within the margin of error any of the polls.

DISTRICT	INCUMBENT	PRE-2021	+0.5 PP
Albany	ALP	5.9%	6.4%
Armadale	ALP	25.2%	25.7%
Balcatta	ALP	8.0%	8.5%
Baldivis	ALP	16.6%	17.1%
Bassendean	ALP	21.6%	22.1%
Bateman	LIB	-7.8%	-7.3%
Belmont	ALP	11.4%	11.9%
Bibra Lake (Willagee)	ALP	17.7%	18.2%
Bicton	ALP	3.6%	4.1%
Bunbury	ALP	10.5%	11.0%
Butler	ALP	20.5%	21.0%
Cannington	ALP	17.6%	18.1%
Carine	LIB	-10.2%	-9.7%
Central Wheatbelt	NAT	-22.2%	-21.7%
Churchlands	LIB	-11.7%	-11.2%
Cockburn	ALP	14.3%	14.8%
Collie-Preston	ALP	14.7%	15.2%
Cottesloe	LIB	N/A	N/A
Darling Range	LIB	-3.5%	-3.0%
Dawesville	LIB	-0.8%	-0.3%
Forrestfield	ALP	9.4%	9.9%
Fremantle	ALP	23.0%	23.5%
Geraldton	LIB	-1.3%	-0.8%
Girrawheen (Mirrabooka)	ALP	23.3%	23.8%
Hillarys	LIB	-0.4%	0.1%
Jandakot	ALP	1.8%	2.3%
Joondalup	ALP	0.0%	0.5%
Kalamunda	ALP	2.3%	2.8%
Kalgoorlie	LIB	-6.2%	-5.7%
Kimberley	ALP	13.1%	13.6%
Kingsley	ALP	1.2%	1.7%
Kwinana	ALP	20.7%	21.2%
Landsdale	ALP	9.6%	10.1%
Mandurah	ALP	18.0%	18.5%
Maylands	ALP	17.9%	18.4%
Midland	ALP	12.8%	13.3%
Mid-West (NW Central + Moore)	NAT	-14.8%	-14.3%
Mindarie (Burns Beach)	ALP	5.4%	5.9%
Morley	ALP	12.3%	12.8%
Mount Lawley	ALP	4.0%	4.5%
Murray-Wellington	ALP	1.7%	2.2%
Nedlands	LIB	-8.0%	-7.5%
Oakford	N/A	N/A	N/A
Perth	ALP	12.6%	13.1%
Pilbara	ALP	2.2%	2.7%
Riverton	LIB	-4.2%	-3.7%
Rockingham	ALP	23.5%	24.0%
Roe	NAT	N/A	N/A
Scarborough	LIB	-5.7%	-5.2%
Secret Harbour (Warnbro)	ALP	23.7%	24.2%
South Perth	LIB	-7.2%	-6.7%
Southern River	ALP	7.9%	8.4%
Swan Hills	ALP	12.1%	12.6%
Thornlie	ALP	15.8%	16.3%
Vasse	LIB	-14.6%	-14.1%
Victoria Park	ALP	16.8%	17.3%
Wanneroo	ALP	8.6%	9.1%
Warren-Blackwood	NAT	-12.8%	-12.3%
West Swan	ALP	18.4%	18.9%

Notes:
Where seats were renamed, their pre-2023 names are provided in parentheses.
No data is provided for Cottesloe, as the most recent by-election gave a 2PP race between the Liberal and Greens. No data is given for Roe as this 2PP race was Nats against Libs.
Darling Range’s data is also based on a by-election.
Geraldton is shown as a Liberal seat even though, as noted above, the member joined the Nationals during this term. With that candidate no longer incumbent and not running in 2025 it is assumed the Liberals will once again be the Coalition party of choice in the seat.
Mid-West is, roughly speaking, an amalgamation of the preceding districts of North West Central and Moore. These were held by Nationals candidates with a 10.1 and 19.5 PP 2PP lead prior to 2021; this has been averaged to 14.3 PP above.
Oakford did not exist prior to 2023, and thus has no data here.

Applying a uniform 0.5 PP swing to Labor across the pre-election 2021 pendulum gives broadly consistent outcomes: Labor 41, Coalition 15 (Liberal 12, Nationals 3).

Of course, the pre-election pendulum doesn’t take into account the boundary changes in the districts (leading to a few of the issues discussed in the associated notes above). To properly calculate this pendulum in anticipation of Saturday’s election, we should actually go to the WAEC’s breakdown of results by polling place (e.g. Albany’s data at https://www.elections.wa.gov.au/elections/state/sgelection#/sg2017/electorate/ALB/results), develop a formula to resolve where these landed on a 2PP basis from the seat’s distribution of preferences data, then add up the place-by-place results within the new district boundaries. That’s an awful lot of work to moderately improve the utility of 8-year-old data which will then be force-fed into a very crude predictive model.

Instead, I’m going to look at the raw, blunt 2PP shift that resulted from Antony Green’s 2023 calculations (aberrant 2021 data and all) and add or subtract these from the pendulum above. So, for example, Albany was held with a 5.9 percentage point margin going into the 2021 election. Mr Green found the 2023 redistribution took Labor in Albany from a 13.7 PP lead post 2021 to 11.0 PP going into 2025.

That’s a drop of 2.7 PP. If we (naively) assume that is directly applicable back through time, we could say the 2023 boundary change drops the 2017 Albany margin from 5.9 PP to 3.2 PP.

This is called the ‘2023 Shift’ in the table below, and is applied to calculate a new pendulum. Where this column fades to white, it represents a shift that narrows the margin of the seat. Where the colour intensifies, it indicates the seat has become more entrenched. Finally, we apply the same 0.5 PP uniform swing based on polling to predict who wins the seat.

Note: the 2023 Shift for Mid-West is derived by averaging the respective -1.7 and -8.5 shifts of North West Central and Moore.

So Where Does That Leave Us?

We’ve applied polling to three different pendula now. In 50 of these seats, the data agrees across all three:

INCUMBENT	2025 -13.7PP	2021 +0.5PP	NEW +0.5PP	AGGREGATE
ALP	-2.7%	6.4%	3.7%
ALP	20.1%	25.7%	24.0%	ALP
ALP	12.1%	8.5%	8.5%	ALP
ALP	22.1%	17.1%	16.0%	ALP
ALP	18.0%	22.1%	22.2%	ALP
LIB	-7.0%	-7.3%	-7.3%	LIB
ALP	15.7%	11.9%	12.1%	ALP
ALP	14.5%	18.2%	19.3%	ALP
ALP	2.5%	4.1%	4.7%	ALP
ALP	8.8%	11.0%	11.0%	ALP
ALP	18.9%	21.0%	21.4%	ALP
ALP	17.0%	18.1%	18.4%	ALP
LIB	-9.8%	-9.7%	-8.3%	LIB
NAT	-23.0%	-21.7%	-20.3%	NAT
LIB	-12.1%	-11.2%	-10.4%	LIB
ALP	14.4%	14.8%	16.2%	ALP
ALP	9.6%	15.2%	15.1%	ALP
LIB	-21.1%	N/A	N/A
LIB	0.4%	-3.0%	-2.4%
LIB	-0.6%	-0.3%	-1.1%	LIB
ALP	9.2%	9.9%	7.3%	ALP
ALP	N/A	23.5%	22.7%
LIB	-4.4%	-0.8%	-3.2%	LIB
ALP	19.7%	23.8%	23.5%	ALP
LIB	5.0%	0.1%	-0.2%
ALP	4.5%	2.3%	0.5%	ALP
ALP	11.8%	0.5%	-0.3%
ALP	0.8%	2.8%	5.5%	ALP
LIB	-2.5%	-5.7%	-6.5%	LIB
ALP	7.7%	13.6%	13.5%	ALP
ALP	3.2%	1.7%	1.7%	ALP
ALP	24.5%	21.2%	24.6%	ALP
ALP	11.9%	10.1%	10.3%	ALP
ALP	10.5%	18.5%	17.5%	ALP
ALP	15.6%	18.4%	18.4%	ALP
ALP	13.2%	13.3%	14.7%	ALP
NAT	-22.3%	-14.3%	-17.8%	NAT
ALP	14.2%	5.9%	6.9%	ALP
ALP	14.2%	12.8%	12.1%	ALP
ALP	8.2%	4.5%	4.8%	ALP
ALP	3.6%	2.2%	2.3%	ALP
LIB	-10.6%	-7.5%	-7.2%	LIB
N/A	14.0%	N/A	N/A
ALP	15.5%	13.1%	13.0%	ALP
ALP	3.9%	2.7%	-0.1%
LIB	-2.8%	-3.7%	-1.8%	LIB
ALP	24.0%	24.0%	24.0%	ALP
NAT	-25.9%	N/A	N/A
LIB	-4.2%	-5.2%	-6.1%	LIB
ALP	17.6%	24.2%	22.1%	ALP
LIB	-3.6%	-6.7%	-6.7%	LIB
ALP	19.4%	8.4%	8.4%	ALP
ALP	13.6%	12.6%	12.8%	ALP
ALP	16.9%	16.3%	16.0%	ALP
LIB	-18.0%	-14.1%	-14.1%	LIB
ALP	14.0%	17.3%	17.2%	ALP
ALP	14.8%	9.1%	9.2%	ALP
NAT	-11.5%	-12.3%	-11.4%	NAT
ALP	21.1%	18.9%	19.5%	ALP

This still gives the Labor party 36 seats, enough to comfortably form a majority government.

The nine undecided seats are:

Albany, which has been a solid Labor district for over two decades. The Shire of Plantagenet is a new addition to the seat with a strong Nationals voting base, but while the Liberals have significantly recovered in the polling data, we’ve not seen similar gains for the Nationals. This’ll surely be a close race, but I’m tipping Labor to hold this one.

Cottesloe, which is a foregone conclusion for the Liberals. This is one of the two seats that the Liberal party held in the onslaught of 2021, and in a year where they make a come-back this is a no-brainer. Cottesloe only appeared as an unclear result because of ‘N/A’ data from pre-2021 – this is due to Labor preforming so poorly here that the 2PP race was between Liberals and the Greens (which the Libs won with a more than 20 PP margin), and there is no real suggestion the Greens could actually unseat the Coalition here.

Darling Range is a genuinely close race. This one really has to be decided by flipping a coin, and on no real basis at all except two of our three calculations turned blue, I’m predicting Darling Range for the Liberals.

Fremantle is under no real threat from the Coalition, and was only flagged for further consideration because on the current pendulum the Coalition doesn’t make it to the 2PP race. This pendulum has Labor’s margin over the Greens at 13.9 in a year when polling suggests the Greens vote is substantially up (more than 50% on last election and a third on 2017). If the Greens are going to win a seat, I’d put my money on Freo (where they did have success in 2009). But this was a by-election with no Coalition participation where the Greens primary vote broke 44%; I don’t see this happening in the 2025 general election and so I’m predicting Labor here.

Hillarys is close on the numbers, with only our outputs from the 2025 pendulum giving a result that’s not within a fraction of a percent of a tie. That said, that’s the pendulum with the aberrant 2021 data baked into it, and I think Hillarys is a perfect example of that behaviour. The district went from a never-red seat to Labor holding over 61% of the primary vote. The numbers say this’ll be a close one, but I think we’ll see a solid return to the Liberals here.

Joondalup is a real flip-flopper of a seat, and Labor won’t hold its 65% primary vote here. Will it turn blue again though? It’s a tossup and probably the closest race in the state this year. But two of our three analyses suggest Labor will hold it, and I backed the Libs on the coin-flip in Darling Range, so for balance let’s call this Labor.

Oakford is the one new district, so there’s not an easily accessible history to draw on. Antony Green’s pendulum gives them a very solid Labor lead, but based on data I think is very, very soft. Oakford’s localities previously sat in the districts of Armadale, Baldivis, Darling Range and Jandakot. Except for Darling Range which is a tossup on our list, these are all Labor seats, and in the first two instances very solidly so. Mr Green’s assessment also has those three districts becoming less Labor-leaning following the distribution, perhaps hinting at some very strong Labor areas being accumulated here. Looking back before the 2023 boundary changes, the Coalition has never won Armadale has (created 1983) or Baldivis (though less impressively; this was formed in 2017 and has only had one member). The Liberals do have a formidable history in Jandakot, but I still have to give the new district to Labor.

Pilbara will be a very close-run race this year. The Liberals haven’t won the seat since Brian Sodeman in 1980, but was taken by the Nationals for one term in 2013. The Nationals have out-polled the Liberals in every election here since they failed to run in 2005, and are the obvious challenger. But they’d have to rely on the swell in Liberal preferences, as the Nationals polling holds steady. Based on very little at all, I’m tipping Labor to hold this seat for now. 

And Roe, a very safe Nationals seat – the safest Coalition seat in the state, and only excluded from the table above by the fact the seat is so supportive of the Coalition that the 2PP race was National vs Liberal. Since 1950 (with a couple of periods where Roe didn’t exist) the Coalition has won every election here (normally by the Nats or their predecessors the Country Party). Which of these parties takes the seat is largely irrelevant to the rest of the state which is only looking to who will give a majority in the Legislative Assembly, but the Nationals regularly outperform Liberals by around 2:1 on primary votes so it seems pretty obvious the Nationals will win this one.

Conclusion

DISTRICT	INCUMBENT	PREDICTION
Albany	ALP	ALP
Armadale	ALP	ALP
Balcatta	ALP	ALP
Baldivis	ALP	ALP
Bassendean	ALP	ALP
Bateman	LIB	LIB
Belmont	ALP	ALP
Bibra Lake	ALP	ALP
Bicton	ALP	ALP
Bunbury	ALP	ALP
Butler	ALP	ALP
Cannington	ALP	ALP
Carine	LIB	LIB
Central Wheatbelt	NAT	NAT
Churchlands	LIB	LIB
Cockburn	ALP	ALP
Collie-Preston	ALP	ALP
Cottesloe	LIB	LIB
Darling Range	LIB	LIB
Dawesville	LIB	LIB
Forrestfield	ALP	ALP
Fremantle	ALP	ALP
Geraldton	LIB	LIB
Girrawheen	ALP	ALP
Hillarys	LIB	LIB
Jandakot	ALP	ALP
Joondalup	ALP	ALP
Kalamunda	ALP	ALP
Kalgoorlie	LIB	LIB
Kimberley	ALP	ALP
Kingsley	ALP	ALP
Kwinana	ALP	ALP
Landsdale	ALP	ALP
Mandurah	ALP	ALP
Maylands	ALP	ALP
Midland	ALP	ALP
Mid-West	NAT	NAT
Mindarie	ALP	ALP
Morley	ALP	ALP
Mount Lawley	ALP	ALP
Murray-Wellington	ALP	ALP
Nedlands	LIB	LIB
Oakford	N/A	ALP
Perth	ALP	ALP
Pilbara	ALP	ALP
Riverton	LIB	LIB
Rockingham	ALP	ALP
Roe	NAT	NAT
Scarborough	LIB	LIB
Secret Harbour	ALP	ALP
South Perth	LIB	LIB
Southern River	ALP	ALP
Swan Hills	ALP	ALP
Thornlie	ALP	ALP
Vasse	LIB	LIB
Victoria Park	ALP	ALP
Wanneroo	ALP	ALP
Warren-Blackwood	NAT	NAT
West Swan	ALP	ALP

That’s a prediction of 41 seats for Labor, 14 for the Liberals and 4 for the Nationals, giving a clear Labor majority in its third term.