In modern Australia, censuses occur every five years. This provides several years between each census for legislation to be changed and for individuals to forget our past battles over collection, storage, control and use of census data that have culminated most recently, in the conflict and dissent surrounding the 2016 census.
In Australia, the issue of privacy is indivisible from that of the Australian census. In point of fact, it was public backlash from the threat to de-anonymise the 1971 and 1976 censuses that gave impetus to Australia's first Commonwealth privacy legislation.
As a result of the protests of the 1970's censuses, the government formed the Australian Law Reform Commission to inquire into the balance between the collection and storage of personal information in Australian censuses and the right of the individual to privacy:
The ALRC received a wide-ranging reference on privacy from the federal Attorney-General in April 1976. At the same time, public controversy arose in relation to certain aspects of the census to be held on 30 June 1976, therefore, the Attorney-General requested that the implications of the census for individual privacy be taken into account in the Commission's general reference. The ALRC released a discussion paper Privacy and the Census (ALRC DP 8) in 1978 and its first report, Privacy and the Census (ALRC Report 12), was tabled in federal Parliament in November 1979. (Quote from ALRC web site.)
The conflict surrounding the 2016 census is rooted in attempts by successive governments to introduce a national identity system and the efforts of Australian citizens to resist these attempts. The most well-known of these campaigns was the Australia Card of the late 1980's through which the Hawke government attempted to introduce a unique identifier which would be used to link together multiple administrative datasets for the purposes of law enforcement. The Howard government had similar plans in mind with their 'Access Card'.
So determined was Labor to introduce a national identity system, it took the Australia Card to a Double Dissolution election, a tactic which ultimately did their efforts no good in the face of the largest nation-wide protests experienced since Vietnam.
In a plan nearly identical to the Australia Card, late last year the Australian Bureau of Statistics took a decision to de-anonymise the 2016 census to enable the creation of a unique key to link together multiple (and undisclosed) administrative datasets.
This was not the first attempt of the Bureau to de-anonymise the census and in fact represents a step in a long history of similar efforts. An Inquiry was carried out in 1997 seeking input on the benefits of a de-anonymised census. Titled, Saving our Census and Preserving our History, the ABS itself was quoted as being clearly against any such de-anonymisation for fear that it would breach the relationship of trust it held with the public and put future censuses at risk (Chapter 3).
3.35 ABS stated that privacy concerns in some overseas countries had adversely affected some censuses. Privacy concerns in the United Kingdom affected both the 1971 and 1991 censuses. Response rates to the 1991 census were so poor that the results are not used as the basis for estimates of the population. The 1981 census is still relied on. In 1971, a public campaign on privacy grounds led to the destruction of that year's census forms in the Netherlands and traditional censuses in the Netherlands are no longer conducted. In West Germany, a substantial campaign of opposition to the 1983 census led to its cancellation because of public concerns about protection of privacy. Following this West Germany made the destruction of census forms mandatory.28
The report went on to detail many potential uses for a de-anonymised census including chasing family members for blood samples to be used in genetic research and assisting family historians of the future (and presumably the companies that now sell access to such records).
5.19 Professor Nicholson told the Committee that for the purpose of genetic research census information would be required well before a 100 year release period. He suggested that information would be required for living persons: We cannot make use of 100-year-old information - particularly obtaining blood samples - because all those people are dead. What we need to do is get to the living members of families that we know are connected through the historical records. That might mean that we go back through this material or other material which would connect them and then we would say, "You are a living descendant of this person."
In 2005, citing the Saving our Census Report, the government amended the Census and Statistics Act to introduce strict liability so that people could be fined for not returning the census form or filling it out incorrectly without the need for the government to prove any motive.
In the same year, the Australian government proposed the idea of using census data to link multiple administrative datasets in a project called the Census Enhancement Project. An independent Privacy Impact Assessment was commissioned which advised against the idea. In the words of Dennis Trewin, Australian Statistician (18 August 2005):
The ABS in 2005 points out that non-statistical or administrative uses would only be possible if radical changes were made to the authorising legislation.
The ABS believes that such changes are extremely unlikely. It would require a Government that would want to use Population Census data in this way, both Houses of Parliament to approve the change in legislation and a compliant Australian Statistician. Changes of this nature would be in clear breach of the United Nations Fundamental Principles of Official Statistics.
The necessary changes to legislation never took place. Despite this, the governance arrangements for oversight of linked data projects put in place in 2010 were quietly removed during 2015 in favour of oversight by the Secretaries and Deputy Secretaries Data Project Boards under the auspices of the Department of Prime Minister and Cabinet and ultimately, the Prime Minister himself (page 31 Public Sector Data Management Project).
Coinciding with this change in oversight of data integration projects, the ABS announced it's intention on November 11, 2015 to consult on the retention of names & addresses: Just five weeks later, on December 18, 2015 the ABS announced it's decision to retain names & addresses.
A census has some interesting statistical requirements, one of which is that it needs at least a 95% completion rate in order to be statistically accurate. Unlike other social research, a census needs extremely high participation rates in order to be usable.
This stringent requirement is one reason former Heads of the ABS have held such a disinclination to de-anonymise the census despite repeated requests from demographers, medical and genealogical researchers who would like to use de-anonymised census data as soon as possible.
In light of the former Statistician, Dennis Trewin's conclusions regarding the collection of census data for use in linking together administrative datasets, it would appear that the recent changes in oversight of data integration projects have been undertaken as an attempt to side-step the legislative and ethical issues identified by former Statisticians. Backing this up, another former Head of the ABS, Bill McLennan who contributed extensively to the Census and Statistics Act 1905, has been outspoken in his belief that the Act as it stands can not be used to compel people to provide name or address as the reason for collection is to create an SLK rather than to create a statistic.
For his part, David Kalisch, the current Head of the ABS claims he has different advice from the Government Solicitor but has not produced that advice for public scrutiny.
Attitudes to risk have clearly changed within the ABS since the 1989 'Save our Census' Report and the 2005 Report by former Head, Dennis Trewin with the current Head, David Kalisch making a decision within the space of a month only eight months before the census to break with history. Unlike previous consultations, this time the ABS did not seek an external review of the risks associated with this option but instead completed the mandatory Privacy Impact Assessment internally.
The issues surrounding the mixing of data collected for different purposes was not made clear during the consultation, nor was the consultation referenced in a single mainstream media story (see media).
When privacy advocates and the general public did begin raising concerns regarding the retention of names and addresses in the weeks leading up to the census, the ABS maintained a deception that no significant changes applied to the 2016 census.
Therefore it can be concluded that at no point either during or after the consultation has the public at large been informed of the significance of the changes to the 2016 census or in many cases that any changes were proposed or decided. This conclusion is reflected in the near universal response to Question 10 in the #CensusFail Survey and has serious legal and political implications.