3.9. The method used to compile census date SLA population estimates is similar to that used for national and State estimates.
3.10. Census counts of usual residents by SLA are adjusted for net undercount using data from the census Post-Enumeration Survey (PES) and demographic analysis.
3.11 The small sample size relative to the large number of SLAs (approximately 1,300 in 1996) places restrictions on the reliability of the PES as a measure of net undercount for SLAs. Consequently, net undercount is estimated using an iterative proportional fitting method (see Appendix 6). Based on the premise that net undercount is related to age, sex and location, it is assumed that differentials for these characteristics at the SLA level reflect differentials at the State level for age and sex, and capital city/balance of State level for location.
3.12. The iterative proportional fitting method uses the following three data sets:
(a) For each SLA, census counts of usual residents compiled by single years of age and sex.
(b) For States, census counts of usual residents by single years of age and sex, adjusted for undercounting.
3.13. Using the data sets in (b) and (c) as marginal controls, the census counts in (a) are adjusted in two stages:
(c) For capital cities/balance of State, census counts of total usual residents by sex, adjusted for undercounting.
- The first stage is to derive census counts of residents in capital cities/balance of State by single years of age and sex, adjusted for net undercount. These are forced to add to both data sets (b) and (c) above.
3.14. Estimates of the number of Australian residents temporarily overseas on Census night, derived from data on residential addresses reported by these residents on returning to Australia after the census date, are added to their respective SLAs. (All persons arriving in Australia, including Australian residents returning, on passenger cards are required to report their intended address which is taken to be the usual residence.) If the census does not occur on 30 June (for example the 1996 Census was held on 6 August) then a further adjustment is made to produce estimates at the nearest 30 June reference date.
3.15. These adjustments take into account any change in the population of each SLA (due to natural increase and net migration between 1 July and 6 August) from the 6 August population estimate. For births and deaths, the actual events registered as occurring between these dates are taken into account. Net migration is estimated based on the relevant proportion of the population who moved within the previous 12 months as recorded in the 1996 Census. Inter-SLA migration is constrained to net interstate migration levels while overseas arrivals was similarly constrained, using overseas migration data. As the Census does not collect data on overseas departures, SLAs are assumed to have the same proportion of State overseas departures as they had of arrivals.
3.16. Finally, the SLA population estimates are constrained to sum to respective State 30 June estimates.
- The second stage is to adjust the census counts for SLAs by age and sex to match the estimates which were derived in the previous step. In the States, but not the two Territories, this is done separately for SLAs in the capital city and balance of State.