With One Nation adjustment 51.76 to ALP
Rolling most recent released 2PP poll by each firm average 52.31 to ALP
The above is what 331 polls looks like condensed into one graph (see rolling roundup for how I dealt with the most recent arrivals). We're here again, Election Day, and I am at most 17 hours from being dragged screaming into the infernal pit of having to find someone to vote 1 for on the Tasmanian Senate ballot paper.
Once again there is so much more I wish I could have done in advance coverage, especially in terms of doing more work on trying to project individual seats (especially those of the non-classic and three-cornered variety) but there is only one of me, and I think my post-election work is more important.
The final Newspoll came out with a 52.5-47.5 2PP in Labor's favour, among a pack of final polls with Freshwater at 51.5 and everyone else between 52 and 53. Although the polls in the campaign were quite diverse (particularly when Labor's lead started picking up) the late polls have been very convergent on the 2PP front despite having a fair range of primary vote estimates. Because of the still mostly poor level of transparency in the industry it is hard to exclude the possibility that some polls are herding, which increases the chance that the average could be out.
If the polls/Newspoll especially are right and 52.5 is where it's at, then in my aggregate a mean outcome for Labor is 79 classic seats - plus any they might pick up from the Greens or other crossbenchers (Fowler their best crossbench prospect), and minus any they might lose to the Greens, One Nation or independents - and none of this stands to be many seats either way. Noting the state poll breakdowns at Bludgertrack (which incredibly include a polled 2PP swing to Labor in WA!) I've taken my seat model and applied half the difference between the state swings and zero to it. On that basis the model has the Coalition favourite to recover Aston (but not very convincingly) and has Labor very wobbly but still just favoured in Bennelong, Gilmore and (without the benefit of any Tasmanian state poll sampling) Lyons. It has Labor a narrow favourite to pick up Sturt and Moore and sees Menzies, Deakin, Leichhardt, Bass and the somewhat neglected Canning as the biggest risks beyond that. (Canning being outer suburban is arguably at less risk than the model says). However there are a lot of seats that the model sees as only a 5-20% chance of falling either way that I am not so sure about at all! If I assume that there is some degree of preference drift and Labor comes down to the high 51s, it is still on for a slightly better than even chance of majority government on about 78 seats (Australian Election Forecasts has 77 - which I rather like just because it would be the same number three elections in a row - off 51.8). But if the 2PP lands below the range of the polling then Labor is probably - but not certainly - falling into minority. Even down towards 50-50 there are still chances for a majority if the seat distribution is fortuitous.
What I find very difficult here is second-guessing which major party seats will fall if there is not a lot of 2PP swing either way. You have to be pretty confident about the strength of the evidence regarding types of seats to pick them, and there is so much data out there in terms of modelled samples from different types of seats, but it's not especially consistent.
Not only the crossbench vagaries but also the vagaries of converting swings to seats always mean estimates like this as to what a given 2PP would mean can be half a dozen seats out. At this election we'll see if the MRP models offered by YouGov and Redbridge have done any better. But even if Labor didn't get a majority off 52.5, it would probably not be far into minority. There are hung parliaments and hung parliaments, and a few seats short is very governable.
I would hope a poll failure on the scale of 2019 is unlikely in view of the progress made in getting more representative samples since then, but I am a little concerned about the number of non-transparent or less-transparent-than-they-were polls in the mix, and the appearance of some herding at the end and also at times through the term. It is very risky to try to second-guess which direction polling error (if there is one) would occur, so for the record a 3-point error in the Coalition's favour would probably see Labor well into minority, but still with a fairly good chance of remaining in government with a seat count around the high 60s. A 3-point error the other way would be a landslide with Labor winning 90 or so seats, and the Coalition losing to a bunch of second-tier teals as well.
Among those polls that break out "independent" and have polled names strictly on the ballot paper, the independent vote is running at around 7.4%, compared to 5.3% in 2022. This is about what I expected given the increase in the number of independents running, including an increase in the number that are Climate 200 funded. It suggests at least that the independent brand is in OK shape - not enough to be sure no-one will lose, but consistent with the idea that the independents generally do OK and their benefit of incumbency helps Climate 200 to throw more resources at the second tier and win some. We will see. As dubious as seat polls often are, the quality of them seen in Green and Independent seats has been much lower this time around (especially we have not had the detailed Redbridge teal seat releases that performed very well in 2022). The Brisbane seats and Macnamara have had a lot of attention this election in terms of modelling and polling, Wills and especially Richmond less so.
The final Newspoll is notable for two other things - firstly the series' highest ever non-majors primary of 33% (in 2022 it underestimated non-majors by a few points too) and secondly the dire net rating of Peter Dutton, whose -28 (32-60) is the second worst election eve netsat for a federal leader ever. Andrew Peacock had -37 in 1990 but still only lost narrowly and even won the 2PP, so perhaps there's hope in there for Dutton yet. (Anthony Albanese's net -10 is the same as Bob Hawke's at the 1990 election.)
I'm posting these late night comments for starters and will add more tomorrow. Tonight I am excited to be working for the Guardian live for the first time and a link will be posted to the coverage I am working on when it starts.
Ipsos Poll: There was an Ipsos released last night which I missed, their only voting intention poll of the term but they have polled in previous terms, with 51-49 to Labor off raw primaries of ALP 28 L-NP 33 Green 12 ON 8 others 14 (undecided 5). I have aggregated this but with a modest weighting as no field dates were published.
Hey Kevin hope you are doing well. What are your plans for this evening? Are you blogging or helping out any of the “old media”?
ReplyDeleteI will be doing live blogging for the Guardian and assisting them with provisional seat calls.
Delete