Australian Bureau of Statistics
2970.0.55.017 - 2001 Census of Population and Housing - Fact Sheet: Confidentiality of Census Output, 2001
Latest ISSUE Released at 11:30 AM (CANBERRA TIME) 03/06/2002 First Issue
|Page tools: Print Page Print All RSS Search this Product|
Confidentiality of Census Output
Security arrangements during collection and processing
Completed census forms will be transferred from the collection centres to the census data processing centre under secure arrangements. Full-time security personnel will be employed to prevent any unauthorised access to the processing centre.
Comprehensive security arrangements are implemented on the ABS computer system. These include the use of regularly changed passwords, access control and audit trails.
Retention of name-identified information
All name identified information from past censuses has been destroyed, once the statistical processing was completed. However, for the 2001 Census, not all name-identified information will be destroyed:
Confidentiality of tabular data
Tables containing cells with very small counts may potentially result in identifiable information. To avoid releasing identifiable information all tables are subjected to two confidentiality processes before release:
These steps are taken to avoid releasing information that may identify particular individuals, families, households or dwellings without impairing the usefulness of the tables.
1. Assessing the size of the table
This process compares the total number of cells in a table to the total population for that table. The total number of cells may include categories such as 'Not stated', 'Not applicable', and 'Inadequately described'.
If the number of cells is the same as, close to, or exceeds the population size, then the table will not be released. This practice avoids the release of tables containing a large proportion of small cells containing identifiable data.
A table of 3 variables cross-classified
This table would result in 460 theoretical cells (i.e. 2 x 5 x 46). The population for this table is 2,942,389 persons. Therefore this table may be released.
Another table of 3 variables cross-classified
This table would result in 5,822,380 theoretical cells (i.e. 2 x 445 x 6,542). The population for this table is 2,942,389 persons. Therefore this table would not be released.
If you are unsure whether a table can be released, please contact ABS Information Consultancy.
2. Introduced Random Error
A technique has been developed to avoid identification of individuals. The confidentiality technique applied by the ABS is to randomly adjust cells with very small values. These adjustments do not impair the value of the table as a whole. The technique allows very large tables, for which there is a strong client demand, to be produced even though they contain numbers of very small cells. It is ABS policy not to release the detailed methodology employed by the ABS to adjust the data.
Tables which have been randomly adjusted will be internally consistent, however comparisons with other tables containing similar data may show minor discrepancies. This is the case for both customised tables and standard products. These small variances can, for the most part, be ignored.
Care should be taken when specifying tables to minimise the number of small cells. No reliance should be placed on small cells. Aside from the effects of introduced random error, possible respondent and processing errors have greatest relative impact on small cells.
More information on random error can be found in the 2001 Census Dictionary (Cat. no. 2901.0), on page 218 in 'Introduced random error'.
These documents will be presented in a new window.
This page last updated 8 December 2006