Way back in August 2011 we raised some questions about how the Australian Census was asking respondents about religious affiliation. Our post suggested that the census seemed to assume that people were members of a religious group, and that a response bias could push people to choose the first bubble on the form (Catholic) even when more and more Australians are reporting being non-religious.
We loved the lively discussion in the comments and on Twitter, but more importantly, people were letting the Australian Bureau of Statistics (ABS) know their concerns about potential bias in the wording of the question. The ABS received 444 submissions about the question — higher than all other categories combined and by a magnitude of three. The ABS’ summary of the submissions:
- the introduction of a filter question asking if the person has a religion with a ‘YES/NO’ response
- the category of ‘No religion’ to be placed at the top of the list - otherwise in ranking order based on output data from the 2011 Census
- revision of the question wording, incorporating new words such as ‘belief’ and ‘practising’
- the inclusion of additional pick boxes for non-Christian religions
- the review of examples of ‘Other’ in the accompanying notes.
This is great - any of these would be welcome changes, with longitudinal annotation of the adjustment, of course. Sadly, the ABS’ recommendation for 2016 is a much more timid change:
The Census is the only source of detailed, small area religious affiliation data. Information on religious affiliation is widely used in the religious community, and by government agencies which provide services complementary to those provided by religious organisations.
The ABS will make some small changes to the list of response categories for common religious groups to reflect the most common responses received in the 2011 Census.
This recommendation is a contradiction. The first paragraph emphasizes the importance of the Census as the only source of religious affiliation data, but failing to address the larger problems with the question makes the results of that single-source wholly untrustworthy.
This has real-life impacts on funding and policy decisions. Religious affiliation data collected from the census is used to plan funding for parochial schools, assigning chaplains in hospitals and the armed services, and determining how to share time in public media. In other words, things that you might notice every day.
The pure volume of submissions on the subject shows that this is an area of great concern for those watching the census. The ABS could have been scientific and statistical about the problem (we would have happily given them our prototype for a trial) but instead they chose consistency over accuracy.
Oh well. Maybe next time.