| Projected 3CP | Projected 2CP | Win Probability |
Click here for full display of Secret Harbour by-election results
Election night commentary
12:20am. The Secret Harbour Early Voting Centre has reported its two-candidate numbers, which closes proceedings for the night. One Nation have a 7.0% margin on both the two-candidate count and my projection, the latter of which I’ll shortly be turning off. The preference flow in favour of Labor has faded a little to 55.2-44.8, now that I’m including all vote types and not just booth votes.
10:20pm. Whatever doubt might have remained has been demolished by the 13,000 votes that have reported on the primary vote — which, contrary to earlier indications, can only be both its early voting and election day results — and it turns out to be One Nation’s best result, at 42.3% of the primary vote, and only an average one for Labor. I’m now projecting a One Nation winning margin of nearly 6%, and I believe that will increase when I have a more definitive preference flow calculation.
10:00pm. The WAEC advises says it “anticipates” that the Secret Harbour Early Voting Centre will report this evening, which it has until midnight to do.
8:49pm. Bit of a lull in the results. We are yet to hear from Coastal Lakes College, from which we can probably expect about 1000 formal votes, or from Secret Harbour Early Voting Centre. The latter, be warned, will report in two stages, starting with about 1000 votes actually cast there on election day. The early votes will come later — probably quite a lot later, and perhaps not tonight. The WAEC reports 12,869 early votes were cast, which “includes mobile polling and Interstate voting”. The former includes 217 votes that have reported their results, which may be the lot, and it also includes the 60 votes cast at the WAEC centre in Perth. Interstate votes I don’t know about, but I’m guessing it’s not many. So presumably there will be upwards of 10,000 Secret Harbour Early Voting Centre early votes in due course. If they break upwards of 53-47 to Labor, we’re looking at a different result.
8:39pm. If you’re enjoying the coverage, please consider rewarding all the hard work that goes into my results system (and everything else) with a donation through the “become a supporter” buttons at the top of the page and the bottom of the post.
8:29pm. Presumably because the ABC has switched off booth-matching and is now reporting the raw TCP as its projection. My own system tries to do something more elaborate with non-booth vote types.
8:22pm. The ABC’s estimate now aligns with my own in giving One Nation a margin of about 3%.
8:19pm. As ever, one remaining wild card: many thousands of early votes still to come. Though a lot of anecdotal evidence to suggest Labor won’t do particularly well out of them.
8:10pm. I note that my system’s projection is very close to the raw TCP result, whereas the ABC earlier projection of the One Nation margin has come down only modestly — 2.7% raw, 2.8% from me, 6.5% at the ABC. Despite what my projection says, Labor wouldn’t be conceding quite yet, but it’s undoubtedly not looking good for them.
8:09pm. There are now five booths in on TCP, and contrary to what I said two updates ago, that 60-40 split in favour of Labor is sticking.
8:03pm. Six booths in now on the primary vote, and the more recent ones have been even worse for Labor than the earlier ones. Nothing further on TCP though.
7:56pm. I suspect that 60% flow of preferences to Labor will fade as more results come in: a big component of it is Port Kennedy Primary School, which was a particularly strong result for the Greens relative to the Liberals.
7:56pm. My system is still calling a One Nation win, but the margin has been slashed, and prima facie the call looks too aggressive for my tastes.
7:52pm. The preference flow is a surprise: 60-40 in favour of Labor. With the next update my system should start using that flow instead of an estimate that was going nearly 55-45 the other way, and there’s a good chance my call of a One Nation win will be retracted.
7:48pm. The WAEC has released the hounds on the TCP count, which it is correctly conducting on a Labor-versus-One Nation basis.
7:47pm. A second batch of postals has been added, the total thereof being up from 1277 to 2597, and it does nothing to change the picture.
7:40pm. There is a consistent picture of a 20% drop in the Labor vote. The One Nation swings are more uneven — the mobile result is the big outlier, probably because it’s attending different places from last time.
7:37pm. It’s all happening now: my 90% ceiling is off and the call for a One Nation win is back on. That’s with a projected margin of 7.7%, which is a lot gentler for Labor than the ABC’s 10.5%. Both of us are applying speculative preference flow estimates, and there are lots of things we might be doing differently in terms of getting swings out of the postals.
7:36pm. I just slapped an arbitrary 90% ceiling on the One Nation probability, just in time for the first polling booth to report from Port Kennedy Primary School, which turns out to be consistent with the impression of One Nation leading on the primary vote and therefore very likely to win.
7:34pm. No, these are “mobile polling” results — I think the WAEC temporarily added them to the wrong row. My point still stands: I will want to see some polling booth results before I have confidence in my projections.
7:31pm. The Western Australian Electoral Commission booth is on. My system is calling it, but I wish it wasn’t — neither of the two results that are in are easy to match to the last election. Uniquely, this booth wasn’t in operation at all then, so I just allocated it 5% of the absent votes cast in 2025 for booth-matching purposes.
7:24pm. My preference flow estimates are more bullish for One Nation than the ABC’s — they’ve got the margin at 6.9%, I’ve got 8.7%.
7:16pm. We finally have results: a batch of 1277 postals, which are hard to booth-match with, so I’d add a note of caution to my projections. Early batches tend to be more conservative but my swing calculation doesn’t accommodate that — Labor had better hope that that’s what’s causing my projection of One Nation winning by an 8.7% margin.
6pm. Polls have closed for Western Australia’s Secret Harbour by-election. As always, the best place to follow the results as they come in is through this site’s second-to-none results facility (see link above), featuring swings, projections and probability estimates, a table recording results and swings at booth level, and – when you press the “activate” button at the bottom of the page – a map display that does likewise. I presume the Western Australian Electoral Commission will be conducting a notional two-candidate count between Labor and One Nation, despite what the relevant table on my results page shows. However, I believe that will be in the realms of mystery until they start publishing results from it, which the WAEC uniquely does not do until it’s convinced it has picked the right two candidates. With 16 candidates, first results could be a while coming.
Everyone’s talking about federal implications, so I’ll have a go at state ones (it’s a while till March 2029, but this was a state by-election): Butler, Mindarie, Landsdale, West Swan, Darling Range, Baldivis are all newly-built outer suburbs where everyone’s mortgaged up to their eyeballs and (I think) whiter than the average, so if One Nation can win Secret Harbour those are on the table too. Maybe Southern River and Jandakot, but they have a big Indian population so maybe less so. Forrestfield isn’t new but is full of tradies and FIFO workers, thanks to being next to the airport. Those are all swing seats (yes, even the one deputy premier Saffioti holds) and in any normal close election the Libs would be chasing them, so if One Nation swipes them instead it’d make things interesting.
William / Kevin / whoever: is there somewhere data about ethnic background / education level / etc by seat level can be found? Stuff like this:
Not on current boundaries – I got that by compiling SA1-level data. I can send you a spreadsheet if you email me.
WB – Any clarification on the preference flow from Lib to ON.
I’ve had a look at the numbers again & I’ll stick with my calc (on election night) that it looks very high to me. 85% ish.
Thanks
nadia,
How do you arrive at 85 percent? My back-of-a-matchbox calculations put it at low to mid sixties.
I concur with Ante Meridian on preference flows. I did a reverse calculation based on the remainder after my guestimates of the preference flows from the other candidates to One Nation, and that resulted in a likely 61% to 62% preference flow from the Liberals.
Nadia – I forgot to mention this before, but I disagree that the DemosAU state poll should be “discarded”.
The primary votes in individual seats can vary significantly compared to the average primary for the outer suburbs. And it is clear that Secret Harbour would be expected to be amongst One Nation’s strongest results in WA (e.g. based on its demographics and its previous voting record).
Not to mention this was a by-election, not a state election, which also affects voting behaviour.
Meanwhile, two seconds later, Hanson’s out there spruiking nuclear power in an electorate which is quite close to where the AUKUS subs will be parked. I’m not sure how well that would have gone down if she had said that before the election instead of waiting until after? I actually don’t know the answer to that. I know there’s discomfort with nuclear in general, and Dutton’s plan didn’t go down well over here in WA, but I don’t know how many of her supporters might be the people who are more open to nuclear and just see it as ‘jobs’ instead of ‘threat’? But Secret Harbour is an electorate that would be in the direct firing line if there was ever a nuclear incident at Garden Island (as would many of my own family who live in the broader Rockingham area). It’s courtesy to release policy information like that before an election, rather than after.
I think ON is cherry picking those seats that they stand a good chance in, and fair enough. However they chickened out in the Qld state seat of Stafford which Labor won albeit by a slim margin, they must have had some good polling to show they wouldn’t poll well which would have put a stopper in their momentum. Instead they left it to their LNP comrades of whom Pauline ( still an LNP tragic at heart) hopes to form a coalition.
There is now a breakdown (results sheets showing first preference votes and full TCP) of the Secret Harbour Early Polling Place Count Results into
“Secret Harbour Early Polling Place – Early polling period only” and
“Secret Harbour Early Polling Place – polling day only”.
12560 and 912 votes respectively for the published total of 13472.
https://www.elections.wa.gov.au/2026-secret-harbour-election-results
Of the 2,251 Liberal preferences accounted for so far (from the early polling station) 1,576 or 70.01% went to One Nation.
There’s been another vote drop.
Count Progress: 78.98%
https://www.elections.wa.gov.au/elections/state/byelection#/2026-Secret-Harbour-By-Election
Results at link. Nothing remarkably new to note.
Absent Votes – 17
Provisional Votes – 848
What are those 848 accepted provisional votes? That feels very high compared to the usual interpretation of name not on roll, already voted, new enrolment, etc. Are they the interstate votes? They don’t feel like the savings provisions votes that got separated out.
And 17 Absent votes?
William Bowe says:
Wednesday, September 9, 2026 at 3:54 pm
The Secret Harbour preference distribution has been published. ON got 59.2% and ALP 40.8% of the preferences when LIB was distributed, which included 79.4% LIB primary votes and 20.6% they received as preferences earlier in the count.
https://www.elections.wa.gov.au/elections/state/byelection#/2026-Secret-Harbour-By-Election
=========
So it’s 79.4% (of the Lib vote flowing to ON).
I was straight out of the blocks at 11.21 pm election night with a rough estimate of 85%.
KB (on the night) suggested it was around the 52.8% mark, this “Topher Field” blogger on Weds 2-Sep said it was around 88%, and AG entered the fray on Thursday 3-Sep with 70.2%. {I dropped a line on AG’s blog to query}
I think I was the closest, so I’ll take this one.
I think the current Federal 2PP needs an adjustment too.
Say if we drop the Lib flow down to 75%, even on last nights YouGov, the ALP would be sitting on 49.7% of the 2PP roughly.
My calc…
ALP Primary (26) plus 25% Lib Primary (4.5) + 90% Green primary (10.8) + 60% Others (8.4) = 49.7
I don’t think the published (respondent allocated) 2PP’s are reflecting what is going on in the country.
She’s overtaking Labor.
I know the “respondent” calculations seem to be the flavour of the month atm, but the actual result in the privacy of the ballot booth, must surely override this.
I think we’re all being led down the garden path on these new “respondent allocated” polls the pollsters are producing.
Although this is just one Federal YouGov, I believe she has overtaken Labor on the 2PP for the first time, with the 8-Sep YouGov.
Alarm bells time.
No Nadia, you’ve read William’s comment incorrectly. 79.4% of the Liberal votes distributed were sourced Liberal votes, 20.6 sourced from other candidates. In total of these votes, 59.8% flowed to One Nation. The figures I published on my blog were flow preferences from tally sheets representing half the vote showing 70% Liberal preferences to One Nation which is consistent with the 59.8% transfer published in the distribution.
OK, fair enough, but focusing on the Lib preference distribution it still suggests that 79.4 of Lib votes (ie: Lib No.1 on the ballot papar), flowed directly to ON – which is a lot higher than the 66% figure Ben Raue calculated after the S.A. election.
Or is that 79.4% figure a misread on my behalf.
I read it as being 79.4% of the Lib votes – at 3PP stage – (had the Libs at No.1 on the ballot, and the remaining 20.6% were added from earlier preference distributions)….. bringing it down to a 59.2% split in favour of ON
Anyway, I don’t wish to make a spectacle of myself given your long experience with all this, but I attest that the current 2PP calculations the newspapers keep publishing are not in the ball park.
Thanks for your comment too. I probably should move it back to your site.
The only reason I’m fluffing around with this Lib to ON distribution is because Secret Harbour is a typical outer suburban seat, and when we had Nepean in Victoria earlier this year, unfortunately Labor didn’t stand a candidate.
I’m just trying to get a grip of the current Federal 2PP calculations.
Note: I didn’t expect the ALP to field a candidate in Nepean just so I could look at the preference flows.
That would be a bit too much.
You’re still reading me incorrectly, and in the exact same way as you were before. When the Liberals were excluded from the count, they had 6018 votes. They received 4777 first preferences, which is where 78.4% comes from, and picked up the rest when other candidates were excluded earlier in the count. This tells us nothing about the flow of Liberal preferences to One Nation or anyone else.
If we want some idea of this, we have to look at what happened next, when the Liberals were excluded and those 6018 votes were distributed between One Nation and Labor. One Nation got only 3549 of them, which is 59%. We can’t infer from this that Liberal preferences split 59-41 to One Nation though, because only 78.4% of this transfer was actual Liberal first preference votes (which is why I pointed the figure out). The data published by the WAEC from the early voting centre suggests it was probably closer to 70-30. What is absolutely clear is that it wasn’t remotely close to 85%.
I’m sat on a ferry in Croatia so this is a bit hard to calculate exactly so
I will approximate.
Say there are 1,000 on exclusion, 800 are Liberal first preferences. Roughly 600 preferences in total flow to One Nation. If the Liberal isolated flow is 80% then that’s 640 preferences, more than actually flowed.
At 70% Liberal flow it is 560 preferences leaving only 40 of 200 other sources votes to flow to match the distribution to One Nation, which I think is low. From the published distribution I’d estimate the real Liberal flow has fallen short of 70%.
“What is absolutely clear is that it wasn’t remotely close to 85%.”
================
OK, Got it. The preference flow is around 70% (from Lib to ON).
Enough speculation from me, and thanks to both of you for your time in clarifying this.
Enjoy Croatia (AG) and enjoy Perth (WB).
I’ve read your blog update (AG) and see you’ve dismantled this “Topher Field” individual.
I don’t know his background, but I’m guessing it’s not politics. I don’t follow his blog/youtube podcast – just heard about it when KB dropped a post on his own site.
70% is the number, so, over & out from me.
Thanks for getting back.
Off the ferry to Korcula and my ride done for the day so I have had the chance to look at the distribution of preferences in detail.
The Liberal Party polled 4,777 votes. During the distribution of preferences, its tally rose to 6,018 votes. When excluded, 3,549 votes transferred to One Nation. That means it is mathematically impossible for Liberal preferences to flow to One Nation at a rate higher than 74.3%. (3549/4777).
My original estimate based on flow data counted out for the Early Voting Centre was that 70.2% of Liberal preferences flowed to One Nation. While I cannot update that number from the distribution of preferences, I strongly suspect the final figure is under 70%.