An overestimation by the ABS of which dwellings actually had people in them has dropped the response rate for the 2016 Census close to the minimum required for the Census to be considered a success.
And there are still concerns in the research community that data from some responses in the half-billion-dollar exercise might be unreliable.
But national and state-level Census 2016 data has been deemed “fit for purpose” by an independent panel, despite privacy fears and a website meltdown that plagued the project.
According to results released on Tuesday, the response rate dropped from the preliminary figure of 96 per cent announced in October to 95.1 per cent. The ABS has previously said the “minimum required response rate” was 93.3 per cent.
The response rate for the 2011 Census was 96.5 per cent.
Australian Statistician David Kalisch said the “quality is comparable to both previous Australian censuses and censuses in other countries, such as New Zealand, Canada and the United Kingdom”.
Canada’s 2016 Census had a response rate of 98 per cent.
The overestimation of occupied dwellings became apparent once the ABS had completed a follow-up survey of tens of thousands of households.
An Independent Assurance Panel was appointed to investigate the accuracy of the Census, but restricted its analysis to only state and national level data.
The panel found that the data was of "comparable quality" to the 2011 and 2006 Censuses. It reported the data was "fit for purpose, useful and usable", and will support the same variety of uses that has been available for previous censuses.
It also found more people reported their age than their date of birth, and there was some reluctance for respondents to commit to a 99-year archiving of their forms, suggesting privacy fears may have spooked some respondents.
Associate Professor Nicholas Biddle, Fellow at the Centre for Aboriginal Economic Policy Research at the Australian National University, said he was reassured by the assessment by the panel around national and state level data.
“But that hasn’t changed my view that before making conclusions about change in socioeconomics or demography that we need to investigate each data item carefully,” he said.
Dr Biddle also highlighted the increase in the number of people who did not respond to some questions.
According to his analysis, 6.9 per cent of the population did not respond to the question on the country of birth, up from 5.6 per cent in 2011.
The question on Aboriginal and Torres Strait Islander origin returned a similar increase in non-responders, up from 4.9 per cent in 2011 to 6 per cent in 2016.
Only the question about religion is considered to be optional by the ABS.
"It could be that the new migrant cohort is less likely to respond, but it also might be that people were increasingly reluctant to do so,” he said.
"The implication though is we do need to be a little concerned about conclusions from individual data items.”
The main surprise from the Census data was the huge increase in the number of people replying that they did not have a religion: 30 per cent of responders, up from 22 per cent in 2011. This was likely due to a reordering of the responses on the form which placed “no religion” first.
The five-year budget for the Census was $470 million dollars.