A poll commissioned ahead of Australia's 2026 census suggests the country's headline religious statistics hinge substantially on how the question is asked. Essential Media surveyed more than 2,000 people using the existing census list-based format and another 2,000-plus using a proposed yes/no format followed by a text box. The former produced 43% selecting 'no religion'; the latter produced 54% The Guardian.

The gap matters because the census, due to be conducted on 11 August 2026, is the dataset governments, faith organisations and funding bodies treat as authoritative. Michael Dove, spokesperson for the campaign that commissioned the poll — Census – Not Religious? Mark No Religion — called it the "gold standard" of data underpinning policy and funding decisions The Guardian. His campaign's position is that the current Australian Bureau of Statistics format overstates religious affiliation and understates the no-religion cohort, a claim it backs by pointing to other surveys on its website, censusnotreligious.org.au, that report higher no-religion shares than the 2021 census figure of 38.9%.

The mechanics of the two formats explain much of the divergence. The existing ABS question presents a list of religions with 'no religion' as one option among many, requiring respondents to actively locate and select it. The alternative tested by Essential Media asks directly whether the respondent has a religion — yes or no — with a text box to specify if yes. Question-order and framing effects of this kind are well documented in survey methodology: leading with a binary choice before offering denominational detail tends to shift response distributions relative to embedding the negative option inside a longer list. The Essential Media findings replicate that pattern almost precisely, with 'no religion' rising 11 percentage points between formats.

Scaled to the population, the yes/no result of 54% corresponds to roughly 2 million adults who would not otherwise register as non-religious under the current format, according to the Guardian's reporting of the poll data The Guardian. That is a meaningful population-level swing for a single question-design choice, and it is precisely the kind of swing that makes census question wording a matter of institutional consequence rather than academic curiosity.

The poll also surfaced demographic texture beneath the topline number. Under the yes/no format, religious affiliation was lower among 18-34 year-olds, at 34%, and higher among university-educated respondents, at 50%. Both patterns are broadly consistent with established generational and educational correlates of secularisation research internationally, though the poll itself does not establish causal mechanisms — it only measures stated affiliation under differing question architectures.

One methodological caveat deserves emphasis for anyone treating this as a direct census proxy: Essential Media surveyed adults only. The census counts children too, typically via responses submitted by an adult on their behalf, and children's religious affiliation figures — often reflecting parental or household default — tend to skew differently than adult self-report. Any extrapolation from the poll to a hypothetical whole-population census outcome under the alternative format should account for that structural difference.

The immediate historical backdrop is that the ABS itself considered altering the religion question after the 2021 census, in which 43.9% of Australians nominated Christianity and 38.9% ticked 'no religion' The Guardian. The ABS subsequently published clarification reports on the religious affiliation question for the 2026 census ABS. That regulatory process has not been uncontested. In June 2026, Catholic bishops wrote to Prime Minister Anthony Albanese expressing serious concerns about proposed changes to the religion identification question, according to reporting at the time The Saturday Paper. Their intervention signals that religious institutions themselves recognise how sensitive question design is to their own headcount, funding formulas and public standing.

What makes this dispute distinctive is that it is not a fight over belief but over instrument design — and both sides have a stake in the outcome. Secularist advocacy groups have an interest in a format that captures what they argue is under-reported non-affiliation; churches, chaplaincy funding, religious education allocations and multicultural service provision all lean on census religion data in ways that make a lower or higher affiliation count materially consequential, not merely symbolic.

The 2026 census will run on the ABS's existing list-based format; no format change of the kind tested in this poll has been adopted for this cycle. The Essential Media results function, then, as an advocacy data point rather than a preview of the actual census outcome — a distinction worth holding onto given how the 54% figure is likely to circulate in coming weeks. Whether the ABS revisits question design for future cycles will depend on how it weighs the methodological critique against institutional pushback from religious bodies already on record with the Prime Minister.