Two rival polls put Australia's 'no religion' count 28 points apart before the 2026 census

Australia counted itself on 11 August 2026, and for the twenty first census in a row it asked what a person's religion is. The question has been on the form since 1911, it is the only optional question of its kind, and in 2021 it produced the finding that 38.9 per cent of Australians have no religion against 43.9 per cent who identified with some form of Christianity. What is new is that two separate commissioned experiments, run within a month of each other, have now put numbers on how much that answer depends on how the question is asked, and they point in opposite directions.

The first was run by Essential Media for the Census Not Religious campaign, a coalition that includes the Rationalist Society of Australia and the Secular Association of NSW. Reported on 20 July 2026, it split respondents into groups of about 2,000 and gave each a different version. Asked in the wording the census actually uses, 43 per cent chose no religion. Asked first whether the person has a religion at all, with a yes or no box before any list of denominations, 54 per cent said no. The campaign's spokesperson, Michael Dove, said the leading form of the census question inflates the religious count.

The second went the other way. On 7 August 2026 the Daily Declaration, a Christian outlet, reported modelling by Compass Polling, a Sydney firm, which randomly assigned 5,012 respondents one of four religion question wordings used in Australian censuses since Federation. The share choosing no religion ranged from 20 per cent to 47 per cent depending on which historical wording they received. The firm's managing partner, Dr Dorian von Freyhold, put the underlying figure at about 26 per cent rather than the 39 per cent recorded in 2021, a gap he translated into roughly 2.9 million Christians misfiled as having no religion. He blamed the primacy effect created when no religion was moved to the top of the list in 2016, and said that while secularisation is real, the instrument measuring it is broken.

Both arguments rest on the same well documented history. Writing in The Conversation on 10 August 2026, Afzal Azeem Chowdhary set it out: 0.4 per cent recorded no religion in 1911, when it had to be written in the margin. In 1971 the form began telling people they could write none, and the figure jumped to 6.7 per cent. In 1991 it became a tick box and reached 12.9 per cent. In 2016 it moved to first position and reached about 30 per cent. By 2021 it was 38.9 per cent, larger than any single Christian denomination. Chowdhary also notes that around a quarter of those choosing no religion in 2021 were born overseas, which complicates any simple story of local disaffiliation.

The wording itself was contested well before the count. The Australian Bureau of Statistics had proposed replacing "what is the person's religion" with a filter asking whether the person has one. Documents released to the Rationalist Society of Australia under freedom of information laws, and published by the society on 13 June 2026, show the Australian Catholic Bishops Conference wrote to Prime Minister Anthony Albanese and to the ABS in March 2024 objecting to the change and to the consultation behind it. The society's account records that one of the bishops' objections, that the new question lacked a yes box, traced back to a draft image the ABS itself had sent the conference on 5 March 2024. In October 2024 the ABS retained the 2021 wording, citing comparability with earlier censuses and insufficient test data. The bureau's 2026 census dictionary entry, dated 21 July 2026, confirms the old wording, ten tick boxes and an other field, and notes that 2026 is the first census to use the 2024 revision of the religious groups classification.

In the week before census night the dispute became a public one. On 6 August 2026 the ABS issued a statement saying it was aware of campaigns aimed at influencing responses, that the question measures affiliation rather than practice, and that people should answer according to their own beliefs. The Catholic Weekly reported on 10 August 2026 that the Archbishop of Sydney, Anthony Fisher, had made the same point from the other side, saying the census is not a test of how closely someone follows religious practice.

What is not known is which estimate, if either, the count will vindicate. Neither the Essential nor the Compass experiment has been peer reviewed, neither firm's full fieldwork dates and question order were published in the reporting examined here, and the two designs are not directly comparable. The ABS has not yet released 2026 religion figures, and when it does, the revised 2024 classification will make the comparison with 2021 less clean than usual.