No still leads in every state in model but leads in Tasmania in recent tiny breakdowns
Estimate includes data to October 1
This is my seventh Voice polling roundup; I expect there will be one more in the final week, but if polls are sparse over the next fortnight I will probably just add those that do arrive in that time to this article. We have only three weeks to go til referendum day and remote voting has started as I write.
There have been any number of items that were supposed by supporters to turn the campaign in Yes's favour - the John Farnham ad (more on that below), publicity of differing Price/Mundine views over Treaty, Nazis and obvious crackpots appearing at small unofficial "No rallies", last weekend's Yes rallies, vast spending and nobody can remember how many campaign launches. For all that the Yes vote is still going down with no point at which any specific thing even appears to have made any difference. There is no reason to believe Yes can get 10% of final support back in the final three weeks when absolutely nothing has even stopped the decline so far and the only way Yes is winning is if there is some strange new kind of polling failure that the world has never seen before.
Many Yes supporters are despondent and preparing to hate the nation and its voters if Australia votes No, especially if it is a heavy defeat. My view is that the many mistakes by the government in pursuing this referendum - above all else holding it mid-term - have made it almost as hard as it possibly could be for semi-engaged voters to find any merit in the proposal. I have found it hard to believe that the Government is that serious about trying to get the Voice passed, as opposed to being able to say that it had done what Indigenous Australians wanted but the resistance had been too much. The Yes campaign has been poor, but it could have been given much better base material to work with. I had a non-exhaustive list of reasons why I think No is winning in the previous edition.
I will highlight a further tension with the specific Voice model that I don't think has had a lot of attention. The theory of the Voice is that Parliament has failed at fixing Indigenous problems through an inability to listen properly. But if this is so, how can we trust Parliament to design the Voice and to understand what structure will make it work fairly? The Voice as proposed leaves all decisions about how the Voice will be constructed to Parliament which has already decided it can't do Indigenous affairs without help because it doesn't know how to listen. The current Parliament would probably do something like accept the Calma-Langton model (two representatives per state/territory, plus one remote representative in five states and one at large Torres Strait Islander representative). That model is in fact severely malapportioned, somewhat like the Senate is, given the great disparity in the Indigenous populations of different states and territories.
I do not intend the above as a knock-down argument for voting No, because there is always the case that any Voice will be better at providing advice to improve outcomes than none at all. And who knows, if there was something wrong with the structure Parliament might get around to fixing it (experience suggests though suggests one should not be optimistic on that score.)
If Australia does vote down the Voice it will not be a unique occurrence. In Chile last year voters heavily rejected a proposed new Constitution, a major feature of which was Indigenous self-determination. Chile has a poll ban in the final two weeks; a couple of pollsters managed to find some way around it but these polls underestimated the No vote anyway, by the equivalent of 3.8% and 9% two-answer preferred. Earlier polls had also generally underestimated No. In the absence of polls, the same space was filled by such nonsense as social media analytics that predicted Yes would win. The result was 38.1-61.9.
Together with all the rest of the poll denying that has marred this Australian campaign, some people have tried to argue that largish crowds at Yes rallies falsified the polls. The truth is that polls aim to gauge the opinions of society at large whereas rallies feature only the highly committed 1% or less at one end of the opinion spectrum (and in this case there are many reasons why Yes supporters are more likely to attend an inner city rally than No voters are). Large rallies and lost electoral causes are old friends who go back to Whitlam, if not earlier.
This week's polls
Four major polls have been added in the last week, each of which showed worse results than the previous poll by the same pollster (in fact there have been 18 consecutive cases of such declines, not counting Pollinate which is discussed below). Firstly there was Essential, switching to a two-week break in Voice polling from the previous four, and finding Yes behind 41-51. Redbridge had Yes behind 38-62 (forced choice) and Newspoll had not much better with Yes behind 36-56. Freshwater had Yes behind 40-60 (undecided excluded, 33-50-17 prior to excluding undecided).
A few things of note here included that this Newspoll found no gender gap (albeit with around 600 respondents of each gender), and that both Newspoll and Redbridge found the youngest age cohort to only be giving Yes a narrow lead (50-41 and 54-46 respectively). Newspoll also tested the urban myth about No voters not showing up (which had its roots in a faulty Newgate question) and found no real difference in how much intending Yes and No voters said they would vote (91% and 90% certain or very likely respectively). Those more likely to indicate doubt were 18-34 year old voters (77%). Essential also reinforced the finding that No voters are firmer than Yes voters (I have an article in Crikey about this), and there are various reports that "soft No" voters tend to be more locked in than they suggest.
Freshwater had a list of reasons that voters who had switched from Yes to No had given for the switch. Voters were allowed to pick multiple answers and the Voice being a distraction from other priorities such as cost of living and housing led the pack on 60%, followed among others by liking the Voice less after knowing more about it (52), finding the Yes case unconvincing (45), and thin-end-of-the-wedge to Treaty or reparations (41). These may be rationalisations rather than real reasons but this was an online poll.
Looking at the simple trend line above, it has Yes on about 40.0% but this is largely a result of the increasing presence of Redbridge (which is the most No-leaning series) in the aggregate. When I adjust for house effects what I've noticed is that the last few months of decline become pretty much linear, something that Simon Jackman also finds in his "Trend" section and I'm seeing the same thing. Using a house-effects adjusted projection, which I am now limiting to the last three months of data, I get Yes on 40.8% and dropping at 0.75%/week. If the more or less linear decline continues Yes now finishes on 38.8, which is a couple of points better than it was two weeks ago. That won't necessarily happen and the polls aren't necessarily spot on so it's still plausible Yes finishes above 40%, but below 35% is still possible too.
State Estimates
There isn't a lot of new state data since the last episode, with Essential and Redbridge only releasing data for the biggest three states and Newspoll presumably saving up for next time when it will have three polls to pool. Both Essential and Redbridge had NSW somewhat surprisingly ahead of Victoria. Essential had the biggest plus score for NSW compared to the national average so far, and had the second worst score for Queensland relative to the national picture that I've seen so far. However an earlier strong NSW result has just fallen out of my aggregate, so all this hasn't made much difference. Here's what I have by my method based on aggregating state/national differences (difference from national in brackets):
Victoria 44.4 (+3.6)
Tasmania 44.0 (+3.2)**
New South Wales 41.7 (+0.9)
South Australia 39.8 (-1.0)
Western Australia 38.3 (-2.5)
Queensland 34.3 (-6.5)
--
Territories Combined 56.8 (+16)**
(** Treat with extra caution!)
Territories Combined, what's this? What it is is that there's a sum to zero constraint on the national total so in theory the aggregated score for the territories can be estimated from the state totals even without any actual territory data. This is an extremely rubbery number and does look a little bit high to me but I will note that in 1999 the Territories finished at +13.6 compared to the national average, and there's potential for Indigenous voters in the NT to push this a lot higher - or not. The NT voting will be fascinating to watch at booth level.
No new data for SA, WA or Tasmania have been added so the uniform-swing target for Yes to score a double majority remains at 51.0. Even if the decline in polling were to stop immediately, polls would still need to be well over three times more wrong than at the 2019 election for Yes to get up.
I should note here the Guardian's tracker which directly aggregates polls of state voting intention, a different approach to mine.
Trollinate?
Media articles about a second survey by Pollinate have been sighted, but nothing resembling an actual polling report or enough detail for me to work out that this is a useable poll, so it is again not included. The second Pollinate survey/poll/whatever found Yes trailing 35-44, compared to 31-39 in their August outing. That is, wait for it, a 0.02% swing to Yes on a two-answer preferred basis! (Actually it could be up to a 1.3% two-answer swing either way depending on the rounding).
What's of most interest is that Pollinate actually surveyed reactions to the much-discussed Yes campaign John Farnham You're The Voice ad. The claim in the media reports is that the ad did not work as 34% of No voters said it had reinforced their No vote while 55% said it had no impact. 66% of undecided voters said the ad had no impact and 16% said it was confusing. This is all of the results I have been able to find.
One problem with drawing any conclusion from this is that when presented with any item of argument for the opposite side, some of the hardest supporters of a given side will always say it made them more firmly convinced of their views whether that is actually true or not. Voters don't always answer that sort of question honestly. The second problem is that ads don't aim to convince everybody - if an ad succeeded in shifting the votes of even 2% of voters that would be unusually good bang for buck; the vast majority of political ads achieve little or nothing. So as evidence that the ad failed, what's been reported of this poll falls well short and is only good for trolling people who thought the ad was going to change everything. That said, we only need to look at the national polls to see the dial not moving after it.
(Some comments of my own about the ad that go beyond my personal dislike of this English song's impact on our popular culture - firstly I thought that the appearance of no fewer than four ex-Prime Ministers (even Rudd!) in the ad would do nothing to connect with ordinary voters who run for the hills if a referendum Yes vote can be connected with Canberra. Secondly the ad portrayed Indigenous Australians pretty much exclusively as either sports stars or recipients of rights, leaving the second ad to ask "Will I be seen beyond the sports field?" because in the first ad, no you weren't.)
I also mention the news.com.au "Great Aussie Debate" survey, a silly piece of incompetent opt-in fluff that has been taken far too seriously by Joe Hildebrand. The takeaway was supposed to be that only 23% of the sample supported the Voice, but in fact respondents were given two No options ("No, there are better ways to solve Indigenous issues" and "I am against the concept") to only one Yes option, and the first of those contained a reason for voting a certain way, so the question was slanted. The results could also have been affected by the question being Q17 in the survey with Q15 having been "Is Australia a racist country?" asked with a bizarre mess of answer options. Hildebrand claimed that the survey's young audience counted for something, but they weren't that young (28.8% under 40). Overall it's the usual - people who read News media sites skew to the right and any opt-in survey of them reflects that and nothing else, no matter what its unscientifically collected sample size.
I expect there will be more polls before too long, so updates will be added until it's time for the final frontier. Unless, that is, anything else comes up before then.
Updates
Morgan (Sep 25)
Same colours as the main chart. This should give a good estimate of what polls expect if undecided voters break heavily to No - around 39%, but note again the disparity between Essential and Newspoll, which between them dominate the graph.
Kevin, any chance of an analytical eye over PMs claim on Insiders Sunday 8th Oct that 1 in 4 haven't made up their mind. Polls seem to me to be in the 8 to 17% "don't know" for the last month. Not sure what basis (other than 25% is about what the Yes needs to find to have any hope of 50% + 4 states) the 1 in 4 claim is being made. Thanks for the analysis in this and all the other posts
ReplyDeleteAs far as I know the basis for it is solely an unreleased internal poll of South Australia by the Labor Party - in other words completely useless. Notably Resolve, which only uses one pass of a question to decide who is "undecided" (and hence gets higher undecided rates than Essential and Newspoll) still had undecided down to 13%. I'll include something about this in this week's article.
Delete