2967.0 - Information Paper: Parliamentary Inquiry into the Treatment of Census Forms: Submission from the Australian Bureau of Statistics , 1997  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 18/07/1997  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All  
Contents >> The Effect of Retention of Census Forms on the Quality, and Hence Value of Data from Future Censuses and Other ABS Collections



In September 1996, shortly after the 1996 Census, AGB McNair was commissioned by ABS to conduct a survey into public attitudes into the issue of retaining census forms. A copy of the final report of that research is included as Attachment 2. The results of that survey, which are discussed in greater detail in Section 4, clearly show substantial public opposition to the retention of census forms, even though this research was conducted straight after a smoothly run census where privacy did not become a significant public issue.

Key findings of the research are:

      • 89% of respondents agreed that "Census forms should be destroyed to protect people's privacy and confidentiality";
      • 67% disagreed that "Census forms should be stored for release in future for research purposes";
      • Between 34% and 45% said they would be less likely to complete a census form if forms were kept for release at some time in the future;
      • Between 38% and 49% said the information on the census form would be less accurate if forms were kept for release; and
      • 73% disagreed that "Researchers should be given access to Census forms including names and addresses".

These findings are consistent with the qualitative evidence obtained from ABS consultations with the community about the census, feedback from Census collectors and other empirical indicators of community attitudes to privacy and confidentiality, such as the increasing proportion of people and households opting to use privacy envelopes.

    The AGB McNair research found that between 34% and 45% of survey respondents said they would be less likely to complete a census form if forms were kept for release at some time in the future. Between 38% and 49% said the information on the census form would be less accurate if forms were kept for release.

    This evidence suggests there is a likelihood of a significant "protest vote" should the current policy on census form destruction be changed. The "protest" could take the form of refusing to complete a form or providing false, inaccurate or even facetious answers. With such strong public opposition a non-response rate of 10% or more is possible. While the Census and Statistics Act, 1905 provides for the compulsory provision of census forms and of accurate data, those provisions are unlikely to be sufficient to ensure an effective census in the face of widespread public opposition.

    It must be remembered that the AGB McNair survey was conducted soon after the 1996 Census, the conduct of which ran very smoothly with no strong privacy campaign in the media. If there were a sustained campaign by civil liberties or privacy groups against the retention of census forms leading up to the next census then the level of non-response could be much higher.

      The AGB McNair research indicates that while a decision to retain census forms might lead to a high level of non-response generally, the level of non-response and the accuracy of answers provided would vary according to geographic location, age, ethnic background, marital status and level of education.

      A different level of non-response between various population groups would make the statistics less reliable. Many detailed analyses using census data could then be biased because the quality of information about some groups would be significantly better or worse than that provided for other groups. It would make interpretation of such results very difficult and could lead to poor or wrong options being chosen for resource allocation or service provision.


      The main purposes of the census are to accurately measure the number of people in Australia on census night to provide a reliable basis to estimate the population of each State and Territory, and to provide information for small geographic areas and for small population groups.

      Lower response rates in the census directly would lead to a less reliable base for the purpose of compiling population estimates. The ABS would then need to put greater reliance on the small survey conducted after the census (the Post Enumeration Survey) to adjust for the undercount and on other less accurate methods of population estimation - these alternatives probably would not be up to the task. Differing levels of response between States and Territories, or between various small geographic areas, would make the adjustment process even more difficult. This means that the accuracy of population estimates would vary across States/Territories, and across other geographic regions.

      [This should not be viewed as only a remote possibility. The lower level of cooperation in both the 1991 UK Census and its Post Enumeration Survey - thought to have been affected by the introduction of a poll tax - led to the 1991 Census count not being of sufficient quality to be used in the rebasing of the population estimates. The Office for National Statistics (ONS) is still having to rely on the 1981 Census as the base for its population estimates, moving them forward over what will be a 20 year period before the next census results are available, using cohort and other demographic analysis techniques and imperfect systems for estimating net migration. The ONS has set up a high level Steering Committee, which includes a representative from the ABS, to oversight the planning of the 2001 Census and the development of techniques and procedures to ensure this highly undesirable outcome does not recur in the 2001 Census.]

      In like manner, lower response rates would mean that less accurate counts would be available on the characteristics of the population. A different level of non-response between States/Territories, between small geographic groups or between small population groups (whether that be migrant groups, low income groups, Indigenous people, people in specific occupations or industries etc) would mean that the statistics at these various levels would be less reliable. Many detailed analyses using census data could then be biased because the quality of information about some groups would be significantly better or worse than that provided for other groups. It would make interpretation of such results very difficult and could lead to poor or wrong options being chosen for resource allocation or service provision.

      The quality of census data very much relies on householders providing accurate responses to the census questions. That is why all census questions are extensively tested to ensure that householders understand them and are willing to respond to them. The provision of inaccurate or false responses would be difficult to identify and, even if identified, would be difficult if not impossible to correct.

      A worse case scenario is that the quality of the census statistics falls to such an extent that it is possible to release only limited data from the census. In that event not only the investment in the census (about $148 million for 1996) would be wasted in part at least, but it would have a devastating effect on the primary uses of the census.


      A High Court decision in the Electoral Case (Attorney-General Cth; Ex rel. McKinlay v. The Commonwealth (1975) 135 C.L.R.1) held that the Constitution required that the population of the various States needed to be ascertained during the life of each ordinary Parliament for the purpose of determining the number of members from each State in the House of Representatives.

      An opinion of the then Law Officers, Attorney-General (Mr Ellicott) and Solicitor-General (Mr Byers), on the High Court decision in February 1976 (Attachment 11) said:

      ".... it necessarily follows that the State's respective populations be reliably determined (emphasis added). For this some method of counting the population such as a periodical census is essential."

      The Law Officers' opinion led to the provisions in the Census and Statistics Act, 1905 which require 5-yearly censuses, and quarterly population estimates.

      On the question of accuracy of population estimates in the context of quinquennial censuses, the Law Officers said:

      "The temporal disparity between quinquennial censuses and triennial elections means that the statistical population estimates which are based on the antecedent census tend to become unreliable and thus to afford ground for a court to hold the number of each State's members in the Representatives is not in fact in proportion to its population."

      The accuracy of State population estimates relies in large measure on the accuracy of the census counts. This accuracy would be directly threatened by reduced levels of cooperation in the census and/or biased responses. There is compelling evidence that the retention of census forms will lead to higher non-response rates in the census, perhaps even of the order of 10% or more if civil liberty and privacy groups mount a substantial privacy campaign. As a consequence, less accurate State and Territory population estimates will result.

      The Law Officers' opinion would suggest that a reduction in the accuracy of State and Territory population estimates could afford ground for the High Court to hold that the number of each State's members in the House of Representatives is not in proportion to its population, as required by the Constitution. The consequences of such a serious possibility would need to be taken into account in any consideration of census form retention.


      Under the States Grants (General Purposes) Act 1994, the Commonwealth Grants Commission bases its allocation of Commonwealth funds to the States/Territories on ABS population estimates. In 1997-98 the amount to be distributed is estimated at $21 billion. If State/Territory population estimates are less accurate because of reduced or biased response in the census, the result is likely to be a misallocation of funds. Underestimating one State's share of the total population by just 0.1% (about 18,000 people) could reduce that State's allocation by $21 million per annum.

      In addition, the Commonwealth Grants Commission relies, in part, on census data to assist in its calculation of the disability factors for the States/Territories. Lower response rates, particularly different levels of response across States/Territories, and inaccurate answers to the census would make such calculations very difficult, if not undermine their accuracy.

      Under the Local Government (Financial Assistance) Act 1995, population estimates at local government area level are also used to help determine the distribution of funds to local government authorities. Significant distortions in the allocation process could result if regional population estimates are inaccurate due to privacy concerns becoming an issue at census time.


      While electoral determinations and allocation of State/Territory and Local Government grants are amongst the more important purposes of the census on a national scale, there are many other important uses of census data. The vast amount of small area and small population group data that the census provides is also used by a host of government and private sector organisations for a variety of purposes, including infrastructure planning, service delivery, and program evaluation.

      In 1993 a Commonwealth Interdepartmental Committee was established to review the census with the specific aim of considering options for reducing its cost. In the process the Committee considered some of the implications of measures that would produce less reliable census data. The Committee identified significant costs to such fundamental activities as designing and managing labour market, education and training policies, and urban planning.

      These included:
      • reduced efficiency of many resource allocation and service delivery programs if occupation and industry data from the census are affected; specific examples of such programs are vocational education and training, workforce planning of the Department of Health and Family Services and economic migration planning by the Department of Immigration and Multicultural Affairs,
      • less reliable data on travel to work affecting decisions on urban planning and infrastructure development,
      • impairment of the only comprehensive source of data on inter and intra-state population movements in Australia, and
      • reduced efficiency in targeting of labour market programs ($1.3 billion budgeted in 1993-94).

      Further, the Committee found that a decline in accuracy of census data would produce a corresponding decline in the reliability of population survey benchmarks used by the ABS in all its household surveys, and hence in the quality of the survey data. The repercussions of this would be felt by the users of a great deal of the ABS's regular social and labour statistics including surveys such as the monthly Labour Force Survey, national surveys of health, income, expenditure, families, time use, and crime victims.

      This view of the importance of accurate census information is widely shared by the Australian community. The release on 16 July 1997 of results of the 1996 Census prompted, amongst others, the following editorial comments on 17 July 1997:

      The Age (Melbourne)

      "Sense and the census

      ....... at a time when Australian politics is increasingly about anxieties generated by a changing national identity, it is reassuring to have reliable data against which claims of politicians can be tested."

      The Courier Mail (Brisbane)

      "The changing picture of Queensland

      The latest census tells us many things we didn't know about ourselves, vividly outlines the challenges of the future, but also demonstrates the capacity of the state and the nation to adapt creatively and calmly to fairly massive demographic and social upheaval."

      The Advertiser (Adelaide)

      "A beautiful set of numbers

      The five yearly Census is one of the best public investments Australia makes. It pays for itself many times over in both the private and the public sector.

      At the everyday human level its findings are engrossing, especially when tracked over time."

      The West Australian (Perth)

      "Census helps us to know ourselves

      Australia's census is an expensive institution which puts every citizen to some inconvenience every five years.

      But it is a valuable device for allowing us to know ourselves. What emerges from the census as a collection of apparently dull statistics in fact paints the most accurate picture possible of contemporary Australian society.

      The figures provide governments with the factual base they need for planning community services.

      Governments at all levels can use the figures to tailor health, welfare, education, transport and other services to identified need.

      But the figures have another purpose: they help us to understand what sort of society we live in and to see where we are heading. This helps to dispel damaging myths on which prejudices are built and to provide a factual basis for community debate......

      The Australian Bureau of Statistics has also produced figures that should get close attention from governments, ......

      The census results have given governments a useful tool for meeting community needs. It is up to governments to make sure it is used properly."


      There is no doubt that the quality and the high level of cooperation that the ABS receives in its statistical collections is based in no small part on the trust the respondents have in the ABS. The cooperation in the census, and indeed most statistical collections, is higher in Australia than in most other countries. This has contributed to the high international reputation of Australia's official statistics.

      The ABS and its predecessor organisations have always attached a high priority to preventing the disclosure of personal information about identifiable individuals. Indeed, the confidentiality provisions of the Census and Statistics Act, 1905 ensure that the ABS cannot release identifiable personal and domestic information.

      ABS is proud of its record in maintaining the confidentiality of information that has been entrusted to it. As the Head of the Privacy Branch in the Privacy Commissioner's Office put it " the ABS is probably the only Commonwealth agency whose assurance of confidentiality means what they say ........ The ABS appears to have an excellent record in relation to this assurance" (Address to the Advisory Council on Australian Archives, 1991). ABS considers this record is fundamental to the success of its continuing operations.

      The trust respondents have in the ABS is based on whether or not the ABS does what it says it will do, especially with respect to confidentiality and security of the data it collects, and also on the perception respondents have in this regard. A change in what has been a fundamental tenet of census taking in Australia, the destruction of census forms, could impact quite unfavourably on this perception. Any diminution in the trust of respondents would impact unfavourably on the quality and level of response ABS receives not just in the census, but in all its statistical collection activities.

      Previous PageNext Page