| Projected 3CP | Projected 2CP | Win Probability |
Click here for full display of Farrer by-election results
End of evening
Writing in advance of last week’s Nepean state by-election in Victoria, I noted that the share of an electorate’s population that was over 18 and had no educational qualifications beyond high school was a highly useful predictor of One Nation support at the South Australian election in March: the specific formula being 0.543 – 0.555x. That formula suggested One Nation would score 26.6% in Nepean and 31.9% in Farrer, whereas the actual results were 24.5% and (on current counting) 39.4%. In both cases One Nation enjoyed the benefit of a forfeit by Labor, whose vote shares at the previous election were and 32.6% and 15.1% respectively. Clearly then Farrer was a stronger result for One Nation than Nepean, and can also be considered above the par of the South Australian result. Two factors likely explain the relatively weak result in Nepean: a less vigorous One Nation campaign and a state Liberal Party in better electoral shape than its counterparts federally and in South Australia.
Another two random observations. First, turnout appears to be on track for around 86%, where the by-election norm is more like 82%. This raises the possibility that One Nation’s surge has enthused previously disengaged voters. Secondly, as I noted during the count, Coalition preferences look to have split about 60-40 in favour of One Nation over Michelle Milthorpe, which is quite a lot stronger for Milthorpe than I would have guessed — about the same as in the independent-versus-One Nation contest for Stuart at the South Australian election, where the independent (Geoff Brock) would seemingly have been more attractive to conservative voters than one who has struggled to shake off the “teal” tag (with how-to-vote cards favouring One Nation in both cases). One possibility is that Liberal and Nationals efforts to tar David Farley with the Labor brush (which “brought 1996 attack ads to a 2026 realignment”, in the estimation of Kos Samaras) had an impact on the voters who remained loyal to them, while failing entirely to entice defectors back into the fold.
Election night commentary
12.15pm. I believe we’re done for the evening, with all booths reported on both the primary and TCP apart from the oddity of the electronic assisted voting result. Late counting will mostly be postals and a small number of provisionals, neither of which are likely to change a One Nation margin that reduced from 9.4% and 7.3% when the big pre-poll booths reported at the very end of the evening.
11.05pm. All but two pre-poll booths are in, recent additions including the Albury pre-poll with nearly 12,000 formal votes, which will knock a point or two off the current raw 59.5% One Nation vote when it reports on two-candidate preferred.
10.24pm. The Griffith pre-poll voting centre becomes the largest booth to report so far with 6582 votes, and it too is a bit below par for One Nation.
10.10pm. The Leeton pre-poll voting centre has provided the first substantial new numbers in over an hour, other than a few special hospital results. Its 2794 votes are a bit above par for Milthorpe and a bit below for One Nation.
8.55pm. The election day booth count is pretty much completed, not counting special hospital and remote mobile results that probably aren’t getting done tonight. However, pre-poll voting centres that collectively accounted for over 38,000 votes are yet to come through, and should in theory do so tonight, though larger centres (Albury pre-poll centre had 12,586 votes) sometimes don’t report until the following day.
8.40pm. My projection and the published TCP result have settled at around 58-42, meaning One Nation have achieved a slightly bigger win over Michelle Milthorpe than Sussan Ley. One Nation are doing at least as well on pre-polls as on election day votes; the postals that are in are stronger for Liberal and weaker for One Nation than the remainder, but about the same on the One Nation-versus-Milthorpe two-candidate count.
8.18pm. David Farley and One Nation are claiming victory. If you’re enjoying my coverage — which among other things provides the only place where you can easily observe the booth results — please consider a tip through the “become a supporter” buttons at the top of the site and the bottom of the post. The results feature in particular involves a lot of work and actually costs money to run, at least in months where the mapping goes over the Google bandwidth limit (as it did very handily in the month of the South Australian election).
8.09pm. This seems a particularly bad result for the Liberals, whose primary vote — barely into double figures — is only slightly shading the Nationals, whose weak showing is less of a surprise. This is never good territory for the Greens, but it may be noted that their vote share is down more than half even in the absence of a Labor candidate. Right-wing small parties are collectively down by around half as One Nation sucks out their oxygen.
8.03pm. Around two thirds of the booths are now in on the primary vote, which includes two of the 12 pre-poll voting centres, both from rural areas. One of which is a little better for One Nation than the local election day booth, the other a little worse. I remain surprised by the evenness of preferences — the estimates I was using at the very start of the count would have allocated about 65% of these preferences, nearly 80% of which are coming from parties of the right, to One Nation. Currently it looks more like 55%.
7.35pm. I’ve just farewelled myself from Ben Raue’s live webcast, which ended with us both concurring that One Nation had gained the seat. My results system had in fact done so nearly half-an-hour earlier, but as noted, it was always my feeling that it was going to go too hard too early for One Nation. Now though we have enough booths in from Albury that it seems that Michelle Milthorpe will at best equal her result there from 2025 against Sussan Ley, and perhaps go backwards slightly, when she needs to be winning these booths big to redress rural booths that are in some cases going against her by upwards of 80-20. The one glimmer for her as I noted was that she seems to be doing better than expected on Liberal and Nationals preferences, but as those parties are only accounting for about 20% of the vote between them, this doesn’t decisively change the overall pictures.
6.30pm. We have a booth in, which is Rankin Springs in the electorate’s north. The good news is that my results system is functioning, the bad news is that it’s been too aggressive too early, so don’t take the “probability” reading too seriously until some real numbers are in.
6pm. Polls have closed for the Farrer by-election. From 6:30pm to 7:30pm I will be appearing on a live webcast hosted by Ben Raue of The Tally Room, so I probably won’t have much to offer in the way of commentary during that period, which will start at around the time the first results are in. We can expect that to happen quite quickly as this is a largely rural electorate with many of the 94 booths (including remote mobile and multiple special hospital booths) dealing with perhaps as few as 150 votes that won’t take long to count. On the other hand, the fact that there are 12 candidates will slow things up at least somewhat.
First, some insight into what my results facility is trying to achieve here. The two-candidate preferred result from the 2025 election was Liberal-versus-Michelle Milthorpe, which few expect to be repeated this time, including the AEC who are conducting a One Nation-versus-Milthorpe count, in line with what is generally expected to be the result. That means there is no historic data to work off in projecting swings in two-party terms. However, my own system is more concerned with projecting off the primary vote, and then filling the gap with an estimate of preference flows. These will be entirely estimated for the early part of the count, and will then switch to the flow indicated by the two-candidate count when enough of it has been conducted.
I am a little wary that my system will be too aggressive in calling it for One Nation based on the first results, which will be from rural areas where One Nation will be especially strong. This becomes troublesome for projection (at least the way I do it) when we have a paradigm shift of a result. The primary vote swing to One Nation in the early booths seems likely to be extremely high, and applying that swing to the final result from 2025 may serve to overestimate them. When larger booths come in from bigger population centres, notably Albury, the One Nation swing may come down, and with it the projected result. In other words, there’s a good chance my system will be calling it for One Nation quicker than I’d like to. If betting markets are on the money, this will ultimately be academic: Betfair is offering $1.10 on a One Nation win, $6.20 for Milthorpe, $10 for the Liberals and $20 for the Nationals.