 |
| Live coverage on election night on Pulse Tasmania - Link will be posted here when known - No paywall! |
TASMANIA 2025 POLLING AGGREGATE (NOT A PREDICTION) Lib 35.0 ALP 30.3 GRN 15.3 IND 14.9 NAT 2.5 SF+F 1.9
IND adjusted for design issues with polling independents
Seat Estimate for this aggregate (total of electorate estimates in brackets) Lib 13-14 (13) ALP 10-12 (12) GRN 5-7 (6) IND 4 (4) NAT 0-1 (0) SFF 0-1 (0)
This article is part of my 2025 Tasmanian election coverage. Click here for link to main guide page including links to seat guides and voting advice.
(18 July: Aggregate has been updated here, with minimal changes.)
This article is not a prediction
Just wanted to make that extra clear! Some people cannot read.
Today's party polling reports
The Mercury today reported an EMRS Liberal Party poll with Liberals on 37 Labor 26 Green 14 IND 19 leaving just 4 for others, which I'm told by Brad Stansfield is based on c. 1000 unique interviews this week and last. On these numbers if accurate I would expect Labor to roughly hold station on 10 or maybe gain one off the absence of JLN, Greens to hold station on five or perhaps gain Braddon, INDs to win the usual four (given that the 19 is likely to be an overestimate) and that leaves about 15 for the Liberals.
Labor gave the media (but not to this stage me) some Pyxis findings about voters saying the government did not deserve re-election (34-53) and apparently some analysis claiming that various polls translated to a large two-party preferred lead for them (which would be interesting to examine since Tasmania has effective optional preferencing at party level). However no voting intentions have been seen from this, and indeed no polling supporting the view that Labor is doing well has emerged to counter the Liberals' narrative that their position is improving. The former finding isn't very useful alone because voters can think a government deserves to be chucked out and also that the Opposition isn't ready, and in these cases voters will often stick with what they know. Also in Hare-Clark where some voters holding this view support parties that run away from attempting to form government when they could be able to, it doesn't necessarily lead to change.
Most seats, does it matter?
Before I start this piece I should just say a few words about the current situation. There have been polls mostly suggesting that the Liberals will win the most seats, but there are also signs that that alone could be irrelevant. In yesterday's leaders debate Dean Winter did not rule out accepting confidence and supply from the Greens without a deal or ministries, and even tried to say he couldn't say no (he can, Governors can be very persuasive but in the end nobody has to be Premier on terms they don't want). He also said he would try to make the Parliament work out of respect for voters. Unless denied, this looks like a strong sign that Labor is preparing if needed to accept confidence and supply from the Greens if offered without conditions, and hence go into government. If Labor doesn't gain seats there would be some embarrassment involved but the line would be yes we didn't do it last time, but this time it's a Budget Emergency so we must accept the support of the Gr**ns. It might not be that simple though - firstly they may need more than just the Greens.
Secondly there is still a pathway if the Liberals do better than their polling for the Liberals to get to 18 seats they can work with - it's just that not a single poll has really supported that pathway so far, though some are close. And it's always possible that if Labor has an in theory route to government but has done poorly they could concede again.
Contra to a since corrected claim in the Conversation today this isn't Europe and a single party winning the most seats has no special status in the process of attempting to form government. There is just a history of leaders saying it does, or setting other markers like most primary votes (which actually did formally matter in the 1950s!). The question is really who can command confidence and supply, however many seats they have that are their own party. In a no-majority situation the incumbent Premier has the call as to whether they wish to test their numbers, which they can do even if they don't have the most seats or there appears to be a deal to anoint someone else.
About this aggregate
Late in the piece in any Tasmanian campaign these days I try to put together an aggregate of what the non-internal polls should be taken as saying collectively. If the polls are broadly accurate, after adjusting them for any obvious reason why they might not be, what should we take them as pointing to? This approach has had a great predictive record down the years (nailing the seat tally in 2004, 2010 and 2021 and being one seat off on two parties in each of 2006, 2018 and 2024) but every year lately I feel that its luck is about to run out in view of limitations of the polling. I am very dearly hoping for one more public poll (at least) so this year's might work better. This particular election so far is not as sparsely polled as 2021, but it is less well polled than 2024. The polls we've had since the election was called variously have little, no or a not very good track record in Tasmania and we're sorely lacking a public campaign period EMRS after they were hired by the Liberals.
I define a public poll as a poll where the commissioning source is not a party and commissions the poll with the clear intention of always releasing it. By this standard there have only been three public polls in recent months, an EMRS in mid-May, the YouGov poll in late June, and the Pulse Media DemosAU July 6-10. Hopefully there will be more. I know YouGov are in the field but am unsure if this is a public poll (I did expect one so maybe tonight or tomorrow). I have put a "V1" on this article in hope that more polling will become available and I'll be putting some new aggregate numbers in a new article if so. If that happens I will link to the new version here.
In the greyer realms there have been a uComms June 10-11 (commissioned by an unknown source and not released but I've seen the results and briefly stated them here) and the first DemosAU that was commissioned by an unknown peak body.
There have also been four three 500-550 vote waves of Liberal-commissioned EMRS samples, but I refuse to aggregate party polling (and after seeing how the federal Liberals' polling went at the federal election, I think people will understand why, though I suspect EMRS are doing the same things they always do and their numbers are more reliable.) I might use seat breakdown data from the party polling if I had the full set, but the numbers that have been released have been cherrypicked. Labor has also referred to party polling, but not to voting intentions, only to numbers regarding voters wanting a change of government and to claims about two-party preferred support (none of which translates in Hare-Clark).
Poll weightings
Because polls are scarce in Tasmania I will usually use several months of polls in an aggregate but with a very low weighting for the older ones. I also don't want any single firm to dominate the aggregate especially if it hasn't been tested at an election here before. I have applied a weighting formula that accounts for the recency of each poll and also my impression of the accuracy of the poll in the Tasmanian context (considering both Tasmanian and national track records and also how much we know about how accurate the poll is, eg we know a lot about EMRS in the Tasmanian situation but not a lot about DemosAU). On this basis the weightings I came up with were:
February EMRS .05
May EMRS .09
June uComms .09
June YouGov .22
June DemosAU .19
July DemosAU .37
The EMRS and uComms polls included Jacqui Lambie Network, who then turned out not to be running. JLN votes in general scatter in preference flows; using Franklin 2024 as a model (because it was a case where their whole ticket was excluded) and with an ambit figure for Nationals based on Nats running ex-JLN candidates I've come up with 25% Labor 23% IND 17% Lib 15% Nat 14% Green 6% SF+F as an estimated redistribution for them. I have also done some redistributions of the Others vote for cases where neither Nationals or Shooters, or one and not the other, were included.
The Independent vote has been overestimated - severely - in polling at the past two elections. It is difficult to say exactly how much by because some of the polls included Independent as a standalone while some lumped Independent with Others. As best I can determine no poll at the last two elections has named all the candidates on the same footing, which I believe is the only way to avoid the problem. My estimate of the average overestimation of the IND vote in five campaign period polls from the previous two elections is 3.8%. I'm not sure the polls are still overestimating what independents will actually get, because there may be a surge to independents in the last weeks of the campaign (DemosAU as the youngest current poll will have an average data age of 11 days on election day) but I think that this is a serious design issue with Tasmanian polling that needs to be addressed, and that the solution is to poll online and list all a party's candidates after its name in brackets.
After redistributing the overestimate proportionally (which may or may not be a good idea) what I get is Liberal 35 Labor 30.3 Green 15.4 IND 14.9 Nat 2.5 SF+F 1.9.
Well I was certainly surprised by the 30.3 for Labor when their most recent public poll was just 24.7! I would think that at the moment many election watchers agree with the narrative that Labor is doing badly and forcing the early election has backfired, and therefore would hold that even the early campaign polls are already totally irrelevant and that the Liberals' internal polling is more accurate.
To explain this number, firstly I believe all the polls are overestimating the independent vote, so Labor's vote is likely to have been suppressed by a point or so in every poll. Secondly with the exception of the two DemosAU polls, all the other polls in the aggregate had Labor well into the thirties - the YouGov outright and the others once I remove JLN from the mix. If Labor has crashed then why did the YouGov taken at the same time as the previous DemosAU have them ahead, and does one or the other have a house effect? (YouGov seat polling performed well in Tasmania at the federal election.) It may well be that things have changed so much through the campaign that the older polls should now be totally ignored, but in that case we don't need an aggregate as we can look at the July DemosAU poll and
see what it is pointing at.
Electorate breakdowns
The next step is to apply electorate breakdowns and here there is unfortunately not much to work with. During the previous term EMRS ran a voting intentions dashboard with rich seat data but it ceased functioning some time late last year; I didn't scrape any of the post-election data from it but in any case anything from 2024 is probably irrelevant anyway. This leaves me with just one YouGov seat by seat sample and two DemosAUs. I could also use the results of the previous election but this is dangerous in view of the large changes in the fields of independents running, particularly in Franklin.
I've merged these seat samples with more weighting to the recent DemosAU and less to the earlier one, and also applied a 2-point adjustment in Franklin where I suspect the Greens are being particularly disadvantaged by Peter George being named as a headline option but not their candidates. As with 2024 there is then a lot of banging clanging and screaming of numbers as I force them to sum to the appropriate levels and what I eventually got was this estimate of how much the polls point to the parties underperforming or overperforming their state vote in each electorate:
Some of these may look and probably are wrong individually, but they are my read of "what the polls say". As applied to the primary vote figures, these then produce this:
Bass: After accounting for independents being likely to be overestimated in the polls, there aren't any fourth forces sticking their head up for a seat here so the aggregate expects the big three to win all the seats. Here Labor would probably beat the Greens, either on outright totals or via an even split between their candidates. The way the Greens would win is if there were strong flows to them from excluded anti-stadium indies, which I suppose is possible. But more likely 3 Labor - which I'm sceptical about firstly because no Michelle O'Byrne and secondly because it's hard to credit that Bass elects only one anti-stadium candidate. It is possible in Bass that enough anti-stadium vote pools with some independent or National to get them over the line (eg if the IND vote is less overestimated in Bass than I think) but at this stage these numbers aren't showing it.
Braddon: Craig Garland hoovers up enough of the independent vote here that he would be expected to win - probably not with quota on primaries but we've seen what he does on preferences last time. The lone Green would be trying to stay ahead of the fourth Liberal and the key question here would be the split in the Liberal ticket. If it was reasonably pro-Jaensch so the Liberals could hold three candidates below quota for a long time they might be able to win the seat, especially as most preferences are bad for the Greens here. However if Labor are well over two quotas the Greens have a preference source, provided that Labor doesn't manage to keep their second and third candidates ahead of them (in which case Labor could win three but it seems very remote). On the aggregate numbers the Greens should win but with a worse Labor primary they may not. (My view is four Liberal is still a real chance here.)
Clark: straightforward on these numbers (Yes there is almost as much IND vote as Green but it will be heavily focused in Kristie Johnston and the rest will scatter).
Franklin: Also appears straightforward on these numbers, though the majors are not that far off an extra seat between them. David O'Byrne would probably be short of quota but not short enough that either of the majors would beat him. Rosalie Woodruff is clear in this projection with the adjustment I made, but not by an enormous amount. The Greens have polled remarkably badly in both YouGov and DemosAU samples in Franklin though much better in the small EMRS subsample the Liberals claimed. While I don't doubt Peter George will knock a hole in the Greens vote I suspect they've struck a few coincidentally really bad samples here based on issues with sampling this electorate.
Lyons: Labor probably just beats the Nationals (or SF+F but I understand their ground game barely exists) on these numbers with the assistance of Greens preferences assuming that the Greens really do that well. However it is possible flows from SF+F and indies could be good enough to get the Nationals home, especially if Greens voters refuse to preference Labor over the stadium.
Anyway, that is how my public-ish polling aggregate stands for now. I hope to have a more data-informed version some time tomorrow!
