The Amazing 2026 South Australian Election: Final Lower House Results And Poll Performance

SA 2026 ALP 34 (+5) Lib 5 (-9) ON 4 (+4) IND 4 (-)
(Changes from pre-election/notional; Labor gained two seats from Liberal during 2022-6 term)

Estimated 2PP ALP 57.89 vs Liberal (+3.3)

Estimated "Shadow 2PP" ALP 58.19 vs One Nation

The 2026 South Australian lower house was remarkable in so many ways.  It makes Queensland 1998 seem almost boring by comparison, except that Queensland 1998 was there first.  Maybe all elections are going to be like this now and this soon will not seem so unusual but if that's so my colleagues and I are going to have a very busy time in the future!

All manner of curious things happened here.  Finally, someone (Lou Nicholson in Finniss) won a state or federal seat from fourth on primaries; hooray we have lived to see it.  Both majors missed the 2CP in Stuart and Mount Gambier in the first such cases since Nicklin 2001.  The Liberal Opposition missed more 2CPs (29) than they made (18) and were outpolled by One Nation (unprecedented) but are still the Opposition.  Worse than that they missed nine 3CPs as well and even managed to finish fifth in Port Adelaide and Black - Black being a seat they won at the previous election!  And so on.  It was obvious this was going to be a very messy election - a little while out I thought how on earth will we ever make a pendulum from THIS - but aspects of it were even more unique than I saw coming.

South Australia 2026: Final Polls Still Have Liberals Third

Average of final week polls ALP 37.1% Liberal 17.6 One Nation 23.3 Greens 11.2 others 10.8

If polls are accurate, Labor wins election easily making several seat gains, with Liberals and One Nation probably winning a few seats each.  Independents could win a few to several seats.

The 2026 SA election has been a weird one for polls.  There's been a very decent amount for a state election, but there were four polls tightly clustered around the start of the formal campaign, Morgan a little later, then nothing til the final days.  Also while the early polls had quite a range of figures, the last-week polls by Fox&Hedgehog, YouGov and DemosAU came out with almost identical primaries.  All have Labor on 37-38%, Liberals 17-19%, and One Nation 21-23%.  Newspoll stood out a bit more with a 40 for Labor and a miserable 16 for the Liberals.  

In my previous article I outlined a way of using the 2022 Legislative Council results to roughly look at where the Liberals and One Nation might make the final two if the polling is accurate, and assuming a swing that is two-thirds proportional and one-third uniform.  I have a few more observations on this model based on the vote share average at the head of this post.  

South Australia 2026: What Can The Right Still Win?

Polling for SA election is very lopsided with right very fragmented

Estimate if Newspoll is correct: approx 1 Liberal and 3 One Nation seats

Off YouGov: approx 4 Liberal and 1-2 One Nation seats

Polling may moderate by election day

------------------------------------------------------------

I've been hoping for enough material to start substantial coverage of the South Australian election and it's finally arrived with polls by Newspoll and YouGov.  Prior to that there had been a string of extremely lopsided polls last year, and a Fox&Hedgehog poll in a similar vein early last week.  The YouGov poll pretty much replicates Fox&Hedgehog's finding that the conservative side is split in two with the Liberals bleeding greatly to One Nation while Labor enjoys a massive lead.  The Newspoll is even worse for the Liberals and could (if it happened) even wipe them out completely.  My estimate is that on average for the Newspoll voting intentions the Liberals would win one seat and One Nation about three, and for the YouGov poll that the Liberals might manage three or four and One Nation just one or two; it's possible there will be about as many (or more) independents as right-party MPs.  However there is a lot still to unfold with where One Nation support goes during the campaign and whether the Liberals can improve.  

What is suggested by the polls so far really aint supposed to happen.  The Malinauskas government is only at the end of its first term, but it is federally dragged, and not even by a first term federal government at that.  John Bannon in 1985 was the last State Premier to get a seat share swing at all in that circumstance.  But these are incredible times in polling and Peter Malinauskas is very popular (with a +40 Newspoll netsat).  At the same time we have what looks like a severe disruption if not a realignment on the right nationally and the SA Libs are a disaster zone.   One has to roll one's eyes repeatedly at news that "Liberal strategists" are hoping for a sympathy vote they don't deserve and trying to argue that a viable opposition is needed.  That worked so well for the similarly hapless outfit that was reduced to two seats in WA 2021.  

Poll Roundup: What Do We Do With A Split Coalition ... Or A Rampant One Nation?

2PP Aggregate 53.5-46.5 to ALP (-1.3 since end of 2025, +0.6 in two weeks)

Shadow-2PP Trend Estimate 54.3-45.7 to ALP vs One Nation.  One Nation has made accelerating gains.

Labor would easily win an election "held now", probably losing several seats

The last few weeks have been remarkably messy ones for Australian politics- and poll-watchers.  Following the Bondi massacre the Opposition pressured the Government to recall parliament as soon as possible to pass measures in response, only to end up wedging itself when it got what it asked for, with the result that the National Party has again broken away.  So should analysts and pollsters still treat these parties as the same entity, and if we don't, what especially do we do in the case of Queensland where Liberal and National party room members run in separate seats under the Liberal National banner?  In the meantime, One Nation has exceeded the ex-Coalition's total in three of the ten polls released so far this year and tied it in two others.  An election right now would make Queensland 1998 look somewhat orderly, with all manner of messy multi-way seats and probably One Nation making fifty or more 2CPs with perhaps something like twenty wins - though this stuff is very hard to model.  So is it time for a Labor vs One Nation "two party preferred" figure as well?

Why Hunter 2025 Can't Be Used To Do Last-Election Labor vs One Nation Preference Flows

With the very rapid rise of One Nation in recent national primary voting intention polling we are starting to see some pollsters offer a national alternative Labor-vs-One-Nation two-party figure.  DemosAU did this in its national Jan 5-6 poll where it found One Nation tied with Labor 50-50 while Labor led the Coalition 52-48, this off primaries of Labor 29 Coalition 23 One Nation 23 (more on that later) Greens 12 others 13.  A newish outfit curiously polling on the same dates, Fox&Hedgehog (founded post the 2025 election by a former Peter Dutton staffer) reported 56-44 to Labor vs One Nation off fairly similar primaries of Labor 29 Coalition 25 One Nation 21 Greens 14 others 11, compared to 53-47 for Labor vs Coalition.  So DemosAU has One Nation two points more competitive than Labor on a head to head with Coalition basis while Fox&Hedgehog has them three points worse.  (I'll add that by my last election preferences 48% 2PP for Coalition is pretty generous on the published DemosAU breakdowns, I get 47.4 as the average for their primaries.)

DemosAU attempted to use last-election preferences by using the flow in the seat of Hunter 2025 (the only federal case ever of a Labor vs One Nation finish) to model Coalition to One Nation flows, by assuming the Greens to One Nation vs Labor flow would be the same as the Greens to Coalition vs Labor flow, and also by assigning flows from Others 50-50 between Labor vs One Nation "As the composition of Others is not known".  (That last bit did strike me as a little curious unless they were also doing the same thing for Labor vs Coalition, given that the 2PP flow from others in 2025 was about 54.7% to Labor).  Fox&Hedgehog simply used respondent preferences.