TWO-ANSWER TREND ESTIMATE YES 47.8 (-2.8 in four weeks)
Aggregated polls have Yes losing in five states and trailing the national average in three (two narrowly)
(Above estimates may be updated if new polls are added in next few weeks)
----
Greetings. I was going to call this article "Welcome to No" but thought of some wrong ways that that could be taken.
Four weeks ago my last Voice update still had Yes very slightly ahead but the lead was not long for this world. Unsurprisingly in this edition Yes is still going downhill rapidly and is now clearly behind in my aggregated estimate. The most recently added polls are Newspoll with a 46-54 result and Resolve with 48-52 (and an even worse 36-42-22 prior to the forced-choice question).
A Yes campaign that started deeply underwater by historic referendum standards is sinking further by the same standards with virtually every poll but many Yes supporters are still hoping that as the campaign ramps up, football final season breeds enthusiasm (I'm not really sure how that works) and greater Yes resources are deployed things will turn around. That would be a new thing in the history of referendums if it happened, but this is an unusual referendum and one conducted in a very different campaigning environment to those in the past.
I've been surprised at quite the speed Yes has declined at since I started writing about Voice polling in April. At first it was not clear that the decline was accelerating.
With no signs of anything stopping the trend so far, there is increasing speculation about what a No vote would mean. One thing that is for sure is that if the Voice does not get up a vast amount of money and energy will have been expended on a campaign for an advisory body that was supposed to encourage Parliament on how to "close the gap" - resources that could have been spent doing so, if anyone spending the money knew how.
There's even been some talk of calling the whole thing off or delaying it til next year because of the perceived consequences of failure but one can only imagine the ridicule that would be poured on a PM who started a referendum believing it would win then ran away because of skanky pollz. What is this, the Commonwealth Games?
Anyway here's the graph again:
Key to colours: Red - Newspoll, Magenta - Resolve, Yellow - Essential, Dark blue - JWS, Light blue - Freshwater, Black - Morgan.
As of 22 July my estimate on a two-answer basis is 47.8% Yes, 52.2% No. This estimate includes corrections for the polls that seem have tended to have better or worse than average results for Yes (Essential, Newspoll and Resolve). Various polls are excluded for reasons explained in earlier editions (click on "voice referendum" tab to see all.
I have the following current state estimates for Yes, with differences from the national picture in brackets. I have made one slight change to the methods stated in the last edition, which is that all Tasmania samples up to six months old will now be included.
VIC 51.1% (+3.3)
TAS 49.3 (+1.5)
NSW 48.9 (+1.1)
SA 47.4 (-0.4)
WA 47.2 (-0.6)
QLD 41.5 (-6.3)
The main changes in the relative picture are that with the monster commissioned March YouGov falling out of the aggregate, SA has gone down and WA has risen. However the result for WA appears a bit high on account of two outlier samples (April Essential and June JWS), one of which is about to also drop out.
I should also note a Freshwater poll of Queensland only (36-50 with 14 undecided) that I haven't aggregated via my methods solely because I'm not aware of it being part of a national sample. Given the time it was taken (late June - early July), adding it would knock about 0.2% off Queensland (which would be neither here nor there really).
Newspoll Gender Break Anomaly
Last week's Newspoll had a very odd result with women (38-49, down from 48-42) supposedly suddenly anti-Voice while men (45-47, up from 38-52) were supposedly now more sympathetic. If that is accurate, this is what that looks like on a two-answer basis:
Indigenous Voter Support Levels
Following the release of the official Yes and No pamphlet there has been an increased interest in the levels of support for the Yes and No cases among Indigenous voters. Tempers are apparently so frayed that I was blocked without warning or comment by Hunter ALP backbencher Dan Repacholi's account on Twitter for nothing more than quote-tweeting a tweet of his that quoted some Yes propaganda on this subject and disagreeing with it. He is the first serving state, federal or territory MP (or I think even local councillor) who I'm aware of blocking me in 11 years, notwithstanding the far greater grief that I have given many others. I guess the size of the block button doesn't matter when ...
The Yes campaign pamphlet claims that the Voice is "backed by over 80% of Aboriginal and Torres Strait Islander people." and doesn't cite a source. The No campaign, also without evidence, claims that "Many Indigenous Australians do not support" what it (all kinds of misleadingly but that would be a subject for a non-polling-specific review) calls "different classes of citizenship through an unknown body that has the full force of the Constitution". "Many" is not quantifiable so I can only note that as an unsubstantiated and uselessly vague claim (how many is "many" if it's still a minority? Would 15% be many? Is it being anecdotally not difficult to find Indigenous voters who will be voting No many?). But the yes side's claim of a specific level of support, for the time being, is also unconvincing.
There have been only two public polls that have either canvassed Indigenous respondents exclusively or produced Indigenous voter breakdowns. Both were commissioned by the Uluru Dialogue, one by Ipsos 20-24 January and one by YouGov 1-21 March. The Ipsos poll found 80% Yes, 10% No, 10% Don't Know from a sample of 300. The YouGov poll found 83% yes (with an undecided option available) from an Indigenous subsample of 732.
These results are inadequate evidence of a support level of over 80% as of July, when the Yes/No pamphlets were compiled, for reasons including the following:
1. They are out of date. In the four months since the YouGov poll was taken general aggregated support for the Voice has fallen by roughly 10%. Perhaps a similar change has occurred among Indigenous voters, perhaps not. The change in the general population suggests fresher polling is required.
2. Both polls had some wording issues. Both polls were taken before the referendum question was unveiled. The Ipsos poll does mention an "alteration" to the Constitution, but then presents answer options of "Yes, in support of a Voice to Parliament for Aboriginal & Torres Strait Islander peoples" and "No, not in support [..]", creating a risk that a respondent might answer based on their general in-principle sentiment about the Voice. The Ipsos poll also falsely states "It is compulsory for Australian citizens to vote “yes” or “no” in a referendum." (There is nothing to stop voters voting informally). The YouGov poll only mentions that the referendum would "establish" the Voice in the Constitution and does not mention that the Constitution would be changed.
3. Remote Indigenous people are hard to sample. Despite stereotypes most Indigenous voters are not remote, but the difficulties of sampling remote Indigenous voters (and remote non-voters because of under-enrolment) in remote parts of the NT are so well known that even NT election seat polls routinely omit majority Indigenous electorates.
4. We do not even know that all respondents were Indigenous. Australia in general uses a three-pronged approach to recognising Indigenous status - descent, identification and community acceptance. Pollsters however accept that a respondent is Indigenous if the respondent says so. Ipsos in particular recruited respondents by asking members of its broader panel a screener question about their Indigenous status. Because respondents receive (albeit small) rewards for completing surveys it is possible that some respondents lied in order to take the survey (gaming screeners is a practice that anecdotally exists but I believe the vast majority of respondents are honest). Others might genuinely but in error believe themselves to be Indigenous.
The Ipsos poll says that the incidence of Indigenous respondents within the panel was 3.8%, which exactly matches official estimates. But it would be a great surprise if Indigenous respondents were in fact proportionally represented on survey panels given point 3 above, so that is consistent with a small degree of over-self-reporting. That said, I would expect that dishonest respondents would if anything dilute the Yes vote.
5. Modest sample size The two polls combined have a sample size of 1032 and a Yes vote of approximately 82.1%. Even treating that as a perfect random sample it is within the in-theory margin of error that Yes could be below 80%. This is a little more so as the effective sample size is 181 in the heavily reweighted Ipsos poll and probably around 625 in the YouGov. That said on a two-answer basis Yes could be expected to do better because of undecided respondents.
Some of these issues can be addressed with fresher polling with more neutral wording, while others are difficult to avoid. Overall while it is very likely that there would be strong majority Indigenous support for the Voice (and the burden is on No campaigners to provide evidence if they believe otherwise), the claim of "over 80%" is distinctly overconfident on the current public polling evidence. There may be private polling but none has been reported.
The material I was blocked by Repacholi's account for questioning also contained a dubious claim about Indigenous leader attitudes to the Voice. It said that "8 out of 11 Aboriginal and Torres Strait Islander leaders are in full support" and linked to this article. But firstly Indigenous MPs are not necessarily Indigenous leaders (although they can be such); they're MPs who happen to be Indigenous. The fact that they are elected by the broader population and not by Indigenous voters per se means that they have no specific mandate on Indigenous matters, their responsibilities are not solely to Indigenous voters, and there is no firm guarantee there will at any given time be any of them in the Parliament. That's one of the reasons the Voice is supposed to be needed after all.
The second problem is the numbers. None of the comments of Senators Price, Thorpe, Liddle or Lambie in the quoted article boil down to "full support" for the Voice, so that is seven not eight on that evidence (6 Labor and one Green). Lambie in particular was quoted as saying "I support the principles of the Voice, but I want more detail. I want to know what a Voice to Parliament will do to close the gap. Right now we're being told a lot of words, but not a lot about the substance behind it and how it might actually work." Lambie has made later comments (eg on Sky), some of which seem to show that she wants the Voice to win (or at least that the referendum being called off would be better than No winning) but she has been very critical of the timing of the referendum and the government's campaign for it and also seems to have mixed feelings about the consequences of even Yes winning because she sees the campaign as very divisive.
As a closing note I've been disappointed by the quality of many of the media "factcheckers" and "annotations" of the Yes and No case pamphlets that have emerged, largely in reaction to disgust about the lack of any requirement for the pamphlets to be factual. For many years now media factchecking has been too often an impotent enterprise that often doesn't check the actual facts being stated, and that for this reason is easily dismissed by those who want to keep on Trumping. I have seen several pamphlet factcheckers that are sloppy, vague, that mistake opinions for facts (especially when they are constitutional expert opinions that other constitutional experts might not necessarily agree with), that rebut claims that weren't made, that fail to indicate whether they even think claims are true or not and that appear to be biased in favour of Yes. I'm considering putting out my own assessment of the Yes and No pamphlets but that is a very large amount of work!
Updates:
Utting WA sample: An Utting Research poll described as a robopoll has produced a much worse result for WA than other pollsters (29-58). This was alongside sensational voting intention findings that had the Liberal Party up 54-46 on the 2PP vote, with swings of 15% 2PP and 20% primary against Labor since a poll just under two months ago when Roger Cook became Premier. However the results of this poll in the 2022 federal election were not good: after initially forecasting doom for the Coalition in several WA key seats its final polls underestimated the damage by an average of 6.1%, though it did correctly predict the fall of Curtin. Given that there is insufficient information about how this poll operates I would suggest treating this result with a Yes vote 10 points lower than any other pollster with much caution. (I have seen a script of a poll operating by SMS texting that may have been this poll unless there was coincidentally another asking the same questions by SMS in the field at the same time.)
Odds Sighted! I have seen betting odds re the Voice for the first time with No at 1.20 and Yes at 4.25. Betting odds favourites don't always win. I understand this market started in early June at Yes 1.70 No 2.05 and wish I had seen it then so I could have commented on how historically silly that was.
Morgan and Labor commissioned polls (28 July): The AFR has reported two commissioned polls, firstly a Morgan for an unstated source and secondly a massive Labor poll with a sample size of over 14000.
The Morgan (taken in what seems to have been late June) was initially reported as finding that 48.2% of Yes voters are firm compared to 27.9% of No voters, but the revised reporting says that in fact 48.2% of the sample was firm Yes, 27.9% firm No and the rest up for grabs. These numbers are wildly contrary to other polls except for those with obvious wording skews and should be disregarded pending full methods and results details.
The Labor poll (the pollster at this stage isn't stated) found a 48-47-5 result, of which just under 40% are hard Yes, about 30% are hard No and 32% are persuadable. This is still considerably different from, eg, SEC Newgate which only found 23% strongly Yes despite having a question wording favourable to Yes - and again it is internal polling which always justifies some caution.
The state breakdowns are reported as having Yes ahead in Vic, Tas and SA and tied in NSW (being behind the national average in NSW is actually bad for Yes but even with this sample size a 1% difference is not statistically significantly behind.) I won't be aggregating the headline result on account of this being a commissioned poll (and with currently unknown wording) but I might include the state results if the full state results, name of pollster and question wordings are released.
No comments:
Post a Comment
The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.