What Can We Really Draw From The Liberal EMRS Poll?

EMRS JUNE 15-17/JUNE 29-JULY 1
FIRST WAVE LIB 32.3 ALP 28.7 GREEN 14 IND 19.2 NAT 1.8 OTHER 3.9
SECOND WAVE LIB 34.5 ALP 28.2 GREEN 13.9 IND 17.8 NAT 2.1 OTHER 3.5

The two waves are statistically more or less identical

Combined they suggest a roughly unchanged parliament 

Today's Mercury saw some numbers from a Liberal Party commissioned EMRS poll taken in two waves of 550 voters ahead of the 2025 election.  I don't include party-commissioned polls in my aggregates (it's bad enough to have to include polls commissioned by unknown forces within Tasmania's perennially bashful brown paper bag "industry groups").  In general parties will make strategic decisions on whether to release polls they have commissioned based on whether they like the results or not, and there is a lot of evidence (cf Freshwater Strategy at the federal election) that internal polls can show parties doing better than they are.  

The Liberal Party might not be delighted with the results of this EMRS polling, but it is much worse for Labor as it shows Labor making no progress towards even being the largest party.  A voter who accepts that will also most likely accept that Labor have sent us to an early election without any real prospect of forming a workable government themselves, and might well want to punish them for that.  But the Liberals are also using the figures to argue that they are in the hunt for four seats in Bass and Braddon and also that Labor might be squeezed to one in Franklin.  (Yep, 3-1-1-2.  It is a set of numbers, I suppose.)

It would be interesting to see what the breakdowns in Bass and Braddon in this two-wave poll are because the Liberal number in Franklin (39.2%, presumably from both polls combined) is 5.8 points above the Liberals's state primary vote.  One might ask when the last time the Liberal Party or even a precursor beat their statewide average by 5.8 points in Franklin and the answer is that the Anti-Socialist Party secured this triumph once in 1909!  In recent decades normally Franklin runs below the state average; it did run 2.2% ahead in 2010 but that's not easy to repeat; Will Hodgmans do not grow on trees.  If the Liberals are getting a 5.1% swing in Franklin and genuinely pushing for four in Bass and Braddon but facing a 3.3% swing against overall then their numbers in the Clark and/or Lyons subsamples would be hideous.

Placing any serious weight on EMRS subsamples of a few hundred votes is a game for mugs at the best of times but the media will never stop playing it ever.  What we most likely have here is simply a randomly off-kilter sample.  Even if we accept the numbers for Franklin at face value (Lib 39.2 ALP 23 Green 16.1 IND 21.6) the IND vote is likely to be somewhat overstated (based on experience from 2024) and even if it isn't it's doubtful that Peter George and David O'Byrne would both win from that anyway.  Perhaps they would if George was only just behind O'Byrne and they ended up more or less even and both short of quota after preferences.  But more likely off those numbers both Labor and independents would leak votes (because there are so many INDs in Franklin) but it could be neither would leak that much since Dean Winter would be somewhere around quota.  Green preferences would probably not flow strongly enough for George to overtake Labor on that.  (The weak flow of Green preferences to George over Labor was a surprise at the federal election, federal Labor's pro-salmon-industry position notwithstanding.)

I've continued to praise and often link to The Mercury's coverage when they do a good job, although they most certainly don't deserve it, but there was sad to say a fair bit wrong with the reporting of this poll and as a result I've decided no linky this time.  For starters it showed no awareness whatsoever of the concept of statistical significance.  Thus, between two waves with a sample size of only 550 each, "the Greens saw a minor 0.1 per cent decrease" (try absolutely trivial), the independent vote "softened" from 19.2 to 17.8 and the Others vote "sank" from 3.9% to 3.5%.  The in-theory margins of error on these and other changes are such that none of the changes are even remotely statistically meaningful; the chance that there was actually no real voting intention change between the two waves is high.  It's this sort of reporting of changes of, in cases, a fraction of a percent that makes many pollsters reluctant to release data to one decimal even though it would be good if more pollsters did so.  The report contained no mention of pollster-supplied margins of error (which would be 4.2% for the samples of 550 and probably around 6.6% for Franklin) despite Australian Press Council guidance that margin of error should be reported or at least taken into account.  

The bigger issue was that the report uncritically reported Liberal Party claims about the poll but there was no independent analysis of the poll and not even a he said she said from the ALP in response.  This smacks of the tendancy that I call "the unhealthy synergy" - commissioned polls that should be reported critically often manage to get compliantly written up by the media they are provided to as free stories.

The poll overall

What I can say about the poll numbers overall is they are within the ballpark of other recent polling, albeit toward the worse end for Labor.   Combined figures of Liberal 33.4 Labor 28.45 Green 13.95 IND 18.5 NAT 1.95 other 3.7 are quite similar to the DemosAU poll with Labor leading 34-26.3 that I analysed here.  If accurate the poll points towards a more or less status quo result where the majors could make small net gains from the non-Greens crossbench (simply because it is less focused in one party with a high degree of internal preference flow) but might not.  I don't expect the independent vote to be as high as 18.5 but even a similar degree of overestimation to 2024 would still put it solidly into the teens (I think something like 14% could happen), and at that level there's strong scope for more than the three independents who won last time to get up, though six combined non-Greens crossbenchers might be a stretch.  The poll also backs in DemosAU in having the Nationals polling badly.  This is, however, party-commissioned polling and only one of the four results I've seen so far was actually a public poll.

The Liberals have been effective in using commissioned polling to shape campaign narratives before, most notably in the 2018 campaign where they used a long string of MediaReach robopolls to paint a picture that only they could win a majority, which over time proved to be true.  It will be interesting to see if anything emerges to counter this one.  So far I'm not seeing this poll's state picture being met with howls of disbelief by election-watchers.  

Update Monday 7th: Seven News has reported that EMRS has projected this poll as 15-16 Liberal, 8-10 ALP, 5 Green and 4-6 Others.  While 15-10 is possible off the overall numbers, in my view it is very unlikely off the overall numbers that the Liberals would gain seats with Labor losing seats.  

Upadate Thursday 10th: The Mercury has reported another wave (the sample size a mere 518) July 6-8 with Liberals 37 Labor 26 Greens 15 IND 18 Nats 3.  I don't like to analyse samples this small in isolation however on these sorts of numbers it is likely the Liberals would increase their seat tally (for instance perhaps winning four in Braddon) while Labor may not.