An increase in omitted responses during the botched 2016 census has been attributed to system failure and heightened privacy concerns.

Despite an independent assurance panel determining the data is “fit-for-purpose” and can be used “with confidence”, there was a rise in unanswered questions and respondents providing false details, such as expletives in the “first name” field.

The Australian Bureau of Statistics’ (ABS) 2016 census was a “digital-first” effort, where the vast majority of Australians completed their forms online, saving taxpayers $100 million in salaries for field workers.

However the plan went awry on the day of the census when millions of people were locked out of the website.

Australians were less likely to provide their date of birth (90 per cent in 2011 and 81 per cent in 2016) and only half of respondents agreed to the National Archives holding their data for the next 99 years, compared to 63 per cent in 2011.

Six per cent of first names and two per cent of surnames were found to be “invalid”, while less than one per cent of Australians replaced names with curse words, titles and other phrases.

The report itself referenced privacy as a possible reason for the increase in missing responses.

Australian National University (ANU) Associate Professor Nicholas Biddle, the deputy director of the Centre for Social Research and Methods, said the surge in non-responses was “undoubtedly in part due to the failure of the online system on the day of the census”.

“But there is also a general decline in response rates to surveys,” he said.

“A larger number of people didn’t give their names or date of birth. It would be hard not to conclude that’s due to concerns around privacy.

“I think the 2016 data is usable and very useful for planning and policy purposes, but analysis of the data needs to be done with care … or we’ll get inaccuracies in measures of change over time.”

The Australian Signals Directorate later allayed fears the system had been hacked, confirming that no data was accessed or extracted.

It was later revealed the problem, which took days to rectify, could have been fixed by switching a router on and off.

The Report on the Quality of 2016 Census Data revealed that 274,000 Australians failed to include their surname, while 209,000 responses were missing the respondents’ first names.

These non-response rates were up to four times greater than those seen in the 2011 census.

Data retention fears

ANU demographer Liz Allen said interest over the census peaked in about February 2016, when politicians expressed concern and the mundane statistical collection became very political.

She said privacy concerns were contextualised in growing public concerns over data retention and government surveillance, with the census becoming “collateral damage”.

“People in the media called for a boycott and sabotage of the data collection,” Ms Allen said.

“Additionally, there was a lot of misinformation published in the media which certainly got reasonable people frightened of the census, but most importantly how the government might use the data.

“The website failure did not help in the matter. But the census remains the only true way to hold government to account, and the ABS and the collection is very much independent of government.”

Ms Allen said the independent quality report suggested some Australians were “more cautious” in their responses.

University of Western Australia’s Dr David Glance, director of the Centre for Software Practice, said some politicians, such as the Greens’ Scott Ludlum, were also actively encouraging people not to provide this data.

“The issue with the quality is that the post-surveys only give you a general idea of the overall numbers of people missed or double counted, but it is likely to be far more lumpy, with specific areas suffering worse than others.

“The system failure would have had a significant impact. I think the fact that it increased the level of effort needed to complete the census means that this would have influenced the answers people gave.

“2016 was a disaster for the census, which was exacerbated by poor leadership and very poor communication. It could have been handled differently and have had minimal impact in transitioning (the census) to online.”