What the polls say

by Dr Paul Read

The Voice to Parliament is the first referendum since 1999, which failed to establish Australia as a republic independent of the Crown. This new referendum, at a cost of $400 million, asks of the Australian people:

A Proposed Law: To alter the Constitution to recognise the First Peoples of Australia by establishing an Aboriginal and Torres Strait Islander Voice. Do you approve this proposed alteration?

One of the unique characteristics of a Referendum is that it requires a double majority to succeed, that is a majority above 50% nationally as well as a majority across states. The problem is that most polls fail to sample enough voters in smaller states like Tasmania and South Australia. One solution is to aggregate results from multiple polls. This was recently attempted by Evershed and Nicholas (26 June, 2023) for The Guardian. Combining results from 10 separate polls demonstrated a massive fall in support from August, 2022 until March, 2023. It fell from around 60% to 50% nationally, the highest support being in NSW (62%) and the lowest in WA (52%). They also found age group influenced support, rising from 40% among those older than 55, up to 84% among 18-34 year-olds.

As of mid-2023, a total of around 60 polls have sampled Australia’s support for The Voice to Parliament, representing a total of almost 100,000 people. Most of these polls are single-issue push polls run by news media, and so prevent deeper analyses on other attitudes and voting habits. There is also some concern that their wording is manipulated, which can have profound effects. For example, the latest Essential Poll of 1125 voters changed its June methodology to include an ‘unsure’ option, prompting a fall in support from 60% to 47% seemingly overnight. Other criticisms of the polls are that they are unbiased slaves to the Murdoch media, and rarely include representation of Aboriginal and Torres Strait islanders themselves. For the referendum to succeed it needs a 50% majority in every state. In the absence of useful sampling in smaller states, some pollsters rely on broad bipartisan support as an alternative marker, where voters of all different stripes offer support.

As will be seen in the following analysis, a majority of these issues are obviated in the original Australian Electoral Study (Cameron & McAllister, 2023), one of the very first polls that examined support for The Voice, and arguably of interest because it takes a baseline temperature of support long before mass media fully weighed in on the debate. As will be seen, those most resistant to The Voice tend to vote for Pauline Hanson’s One Nation and the Nationals. Whilst Pauline Hanson’s One Nation embraces a slew of policies that are ‘tough minded’ and punitive, other analyses suggest that Nationals voters can adopt progressive policies depending on context. Meanwhile, Greens, Independents and Labor voters tend to support the ‘Yes’ campaign. This article also examines resistance to the campaign and identifies strategies for engagement with the ‘No’ camp.

pseph·ology (noun) the study of elections and voting statistics
Cambridge Dictionary

The original Australian Electoral Study (Cameron & McAllister, 2023) sampled 2508 people from around the country, asking whether they would support a referendum to recognise Indigenous Australians in the Constitution. This paper looks at the results and trends that emerge from this question using both the original sample (which was skewed towards an older cohort and so possibly more conservative) as well as a randomly selected subsample of 1497 people that more closely match Australia’s age structure. A proper survey reflecting the entire nation would have required a sample size of at least 5000 people so results should be viewed with caution. Only one survey has achieved this — the YouGov survey of 15,060 people in March 2023. The following analyses examine the original results from the Australian Electoral Study (2022) contextualised against results from the largest and more recent YouGov survey commissioned by the authors of the Uluru Statement from the Heart.
In the original 2022 AES survey, the first question was how much people support the change overall. As can be seen in Figure 1, fully 85.7% of the sample either supported (39.4%) it or strongly (46.3%) supported it. Responses were recorded on a four-point Likert scale where 1 denoted ‘strong support, 2 was ‘support’, 3 was ‘opposed’ and 4 was ‘strongly opposed’. This is almost identical to the full sample’s result, 81.8%. By contrast, the more recent YouGov survey puts support at 51%, opposition at 34% and those ‘undecided’ at 15%.

Figure 1. Number of people endorsing a referendum to recognise Indigenous Australians

The support by state is shown for AES (2022) in Table 1. Again, sample sizes were too small for Tasmania, Northern Territory and the ACT, but the trend overall suggested each state was broadly in support of the Referendum. These findings, although the sample size was small, suggested The Voice would succeed in each state. The larger and more recent YouGov results show the same trends across states but at smaller rates of support that tend to match only the values for ‘strong support’. They diverge only in the case of Tasmania, where the larger sample predicts far less support than the original AES sample. It falls from 66% strong support to 50% support. In the absence of Tasmania, excluded as an outlier, the correlation between the two samples across states is r2 = .90 (p < .001).

Among Indigenous Australians themselves, there were only nine males and 13 females recorded in the AES sample. The rate of 77% support in the tiny AES sample roughly matches the more recent 83% support in YouGov, which was based on the largest sample of Indigenous Australians to date, a total of 732 people. When the Indigenous AES sample was combined into both genders 18% ‘opposed’ it and 5% ‘strongly opposed’ it. To increase the sample size the original sample of 2508 was checked and the same trends emerge in a slightly expanded sample of Indigenous Australians — 59% ‘strongly support’ it, 22% ‘support it, 15% ‘opposed’ it and 3.5% ‘strongly opposed’ it.
When tracked across five-year age increments in the AES, all ages are largely supportive until the age of 53, at which point older respondents become slightly resistant to supporting the Referendum. As to gender, 82% of men are supportive and 89% of women are supportive. This also matches the trend in the YouGov survey.

Figure 2. Endorsement of Referendum by five-year age categories

Income, education, and voting preferences for 2022 also have effects on support for the Referendum, at least as available in the AES (unavailable in the YouGov survey). Those who have a university education (56.7%) are significantly more likely to endorse the Referendum (1.6) versus those without (1.8), yet both average out as supportive to some degree (p < .001). Support grows with income, ranging from 2.0 for those below $30,000, through 1.8 thereafter up to $150,000, and then reflecting stronger support at 1.5 above that (r2 = .61, p < .001). Where sample sizes were large enough to make an appropriate assessment on actual voting from 2022, support was strongest among Greens voters (1.4), followed by Independents and Labor (both 1.6), then tends towards opposition among Liberal voters (2.1), followed by Liberal Democrats (2.3), Nationals (2.4), and then Pauline Hanson’s One Nation (2.6). A series of Spearman rank correlations (as appropriate for Likert scales of this nature) sought to identify correlates to opposition in the AES sample. The strongest results are tabulated below, first for opposition and then for support.

Table 2. Correlates to opposing The Voice, including Spearman’s Rank, by importance.

Issue Correlation Significance
Turn back asylum seekers .47 <.001
Death penalty for murder .38 <.001
Defence .28 <.001
National security .23 <.001
Tougher sentencing .21 <.001
Police and law enforcement .17 <.001
Taxation .14 <.001
Business and industry .11 <.001
Economic management .11 <.001

Table 3. Correlates to supporting The Voice, including Spearman’s Rank, by importance.

Issue Correlation Significance
Aboriginal land rights .69 <.001
Government support for indigenous .66 <.001
Equal opportunity for women .49 <.001
Equal opportunity for migrants .47 <.001
Global warming .44 <.001
Environment .39 <.001
Unemployment benefits .32 <.001
NDIS .31 <.001
Education .31 <.001
Refugees and asylum seekers .29 <.001
Voluntary euthenasia .22 <.001
Decriminalise cannabis .21 <.001
Old age pensions .21 <.001
Public transport .20 <.001
Ordinary people need more power .19 <.001
Affirmative action for women .18 <.001
Health and Medicare .14 <.001

As can be seen the hot button issues that tend to correlate with The Voice include a set of (likely intercorrelated) factors reflecting conservative views in opposition and factors reflecting progressive policies in support. For context, these significantly emerge across a total of 41 issues covered by the AES, ranging from defence to gender politics. So, these 22 items emerge from a broader set of themes as significant. Interestingly, the conservative factors list three items related to economic management and six items related to what might be seen as a ‘tough-minded’ or even a punitive stance regarding law and order. By contrast the 13 supportive items are more focused on areas where government is supposed to intervene — transport, education, health, pensions, disability services, unemployment. They are more civic in nature. Outside of such issues are progressive views on gender, euthanasia, cannabis and climate.

The final analysis looks at how Anthony Albanese is viewed by respondents, split by their support versus opposition to The Voice. This is because he has been largely seen as the champion of the cause and roundly targeted for criticism by the Opposition. Across nine personality descriptives, supporters viewed Anthony Albanese significantly more favourably than those who opposed it (based on independent sample t tests applied to mean scores, p < .001). Results based on raw population frequencies for endorsement of ‘extremely well’ and ‘quite well’ are summed in Table 3 for the supporting group, the opposing group, and the total sample for context.

Table 4. Attitudes towards the Prime Minister according to those who support vs oppose The Voice

Issue Support (%) Oppose (%) Total
(%)
Inspiring 52 22 48
Knowledgeable 66 39 61
Strong leader 65 35 61
Trustworthy 70 35 65
Competent 72 40 67
Intelligent 74 45 70
Honest 74 45 70
Sensible 79 50 75
Compassionate 81 59 78

In addition to the question on constitutional reform there are two other questions in the AES that ask about Indigenous issues, which can be used to check the findings above. The questions ask whether the respondent agrees that government help for Indigenous Australians and Aboriginal land rights have gone too far, not far enough or is about right. This is answered on a five-point Likert scale where 1 denotes ‘gone much too far’, 2 was ‘gone too far’, 3 was ‘about right’, 4 was ‘not gone far enough’, 5 was ‘not nearly gone far enough’. National endorsement of more action on these items (summing 4 and 5) cumulatively represents 48% for land rights and 49% for government support, both much lower than the 86% support for The Voice in this sample. Split by state, there is convergent validity across the questions because they correlate with responses on the Referendum, r2 = .72 (p < .05) for land rights and r2 = .57 (p < .05) for government support. Aboriginal and Torres Strait Islanders themselves are more likely to endorse more action on land rights (53%) and government support (55%) than the general population (48% and 49%), and yet less likely to support The Voice (77% Indigenous versus 86% generally). Whilst these questions offer convergent validity across questions, the much higher endorsement of The Voice compared to action on land rights and government support across both Indigenous and non-Indigenous Australians might suggest the referendum question was phrased in a way almost guaranteed to garner support. Framing in the mind of the respondent is important because the obverse of recognising Indigenous Australians is not easily defensible — outright non-recognition, which is implied, is morally hard to endorse at the extreme.

Table 5. Endorsement of Indigenous issues by state and Indigeneity (%)

Land Rights Government Support Referendum
New South Wales 48 51 86
Victoria 52 55 86
Queensland 41 37 82
South Australia 36 35 81
Western Australia 44 49 88
Tasmania 66 60 100
Northern Territory 40 40 90
Australian Capital Territory 60 65 94
Indigenous 53 55 77

If The Voice is to achieve majority endorsement in every state, there is one final analysis that might prove helpful focused on the predominant media engagement of the opposing groups. The results are outlined in Table 6, where extreme media engagement is rank ordered for those who ‘strongly oppose’ versus ‘oppose’ the Referendum. This highlights where activism might prove most useful, perhaps targeting the less rusted-on group in the softer opposition camp rather than those who ‘strongly oppose’. This list likely reflects the key campaign strategies of the No Campaign, or at least hints at the most effective strategies for media engagement for the Yes Campaign.

Table 6. How to reach the opposing camp — engagement with media by rank ordering

Measure ‘Oppose’ ‘Strongly Oppose’
Political activist approach by any means 42 51
Mass media websites 50 43
Political activist approach by mail 39 35
AEC website 35 31
Internet 36 30
Voting advice website 24 28
Television 29 26
Radio 12 25
Political activist approach by text 22 24
Facebook, Twitter and Whatsapp 22 16
Campaign websites 21 16
Newspapers 11 14
Political activist approach by telephone 11 10
Political activist approach face to face 6 6
Political activist approach by email 13 6
Federal parliament websites 10 2
Election blogs 12 0

Conclusion

The question was: “If a referendum were held to recognise Indigenous Australians in the Constitution would you support or oppose such a change to the Constitution?” Fully 86% of the population supported the Referendum based on this question, although only 46% strongly supported it. Among those states with a large enough sample size there was support in NSW, Victoria, Queensland and Western Australia, whereas a small sample of Indigenous Australians were less supportive. Being younger or female garnered more support, as did a university education and greater income. Voting Greens, Independents and Labor was associated with support, with all other major parties being resistant. Positive ratings of the Prime Minister were also associated with support. Whilst supporters are generally more concerned with civic issues of education, transport, pensions and benefits, they also endorse progressive policies on euthanasia, the environment, cannabis and gender. Conversely, those in opposition to The Voice tend to endorse issues of economic management, taxation, and more punitive measures aimed at capital punishment, asylum seekers, defence and tougher sentencing. As to media engagement, the group most resistant seem to be more engaged, if not influenced by, in order, direct approaches by political activists and mass media websites, mailouts, electoral and voting websites, and television. Finally, whilst the AES suggests broad-based support across all states, and that a Yes campaign would win, the sample size is simply too small to put too great an emphasis on its outcomes.

References

Cameron, Sarah & McAllister, Ian (2023) Australian Election Study, 1987-2022 Trends, doi:10.26193/HPA0BY, ADA Dataverse, V1
Evershed, Nick & Nicholas, Josh (2023) Indigenous voice to parliament referendum tracker: how many people support or oppose. The Guardian, 26 June, 2023.