A1.1. Estimates of the census net undercount, as derived from the Census Post-Enumeration Survey, are subject to adjustments based on demographic analysis ('demographic adjustments'). These demographic adjustments are made to the estimates of net undercount at the Australian level, on the basis of investigations of the population numbers and the sex ratio (number of males divided by number of females) of the population estimates by 5 year and single year age groups. The estimated resident population derived through using the PES estimates of net undercount is compared with:
- Demographic estimates, derived from the National Demographic Databank. This database is maintained according to administrative data, covering births, deaths and permanent and long-term arrivals and departures. The data are seen as representing total population well up to about age 30, after which there are some concerns about pre-1970 migration data. Any inaccuracies in category jumping estimates could cause a gradual erosion of a cohort's reliability in the Databank. Sex ratios however appear to become less reliable for ages greater than about 22.
- Medicare enrolment numbers. These numbers are seen as a good basis for calculating sex ratios, but are seen as less reliable for population totals (especially for later age groups where deaths often have not been removed from the Medicare register).
A1.2. There appeared to be a clear pattern in 1996 in which the numbers of those people aged 30, 40, 50, 60 and 70 were slightly high, and the numbers of people aged 31, 41, 51, 61 and 71 were slightly low in comparison with the independent sources. As a result of this 'age heaping', the population in these pairs of consecutive ages were combined and then redistributed according to the Demographic Databank age distribution.
A1.3. In addition, the unadjusted PES under-enumeration rates for each sex by 5 year age grouping were compared with each other and the equivalent rates in 1986 and 1991. A table of these unadjusted rates is included below:
- The Estimated Resident Population based on the previous census. Note however that the Estimated Resident Population based on the previous census is sometimes correlated with the other two sources due to adjustments made on the previous census.
National Demographic Data Bank
A1.1: Census Post-Enumeration Survey Net Undercount
* Undercount rates in these cells were adjusted in 1991 for internal consistency with other age groups (effect not shown)
** Undercount rates in these cells were adjusted in 1991 for consistency with other demographic data in 1991 (effect not shown)
*** Undercount rates in these cells were adjusted in 1996 for internal consistency with other age groups and/or consistency with other demographic data in 1996 (effect not shown)
A1.4. In the 1960s, before the Post-Enumeration Survey (PES) was fully operational, a computer file was created by H.P.Brown of the Australian National University (ANU) to enable an assessment of census under-counting and age mis-statement by demographic analysis. Called the ANU Demographic Databank, this file contained a time-series of national demographic events (births, deaths and overseas arrivals and departures) and census counts of population for all years back to 1921. The data, which was by single years of age and sex, was obtained from the ABS, as was additional data used by A.R.Hall of the ANU to update the file for the period 30 June 1966 to 30 June 1976. (A. R. Hall, 'Australian Demography Data Bank, Vol.1 (1968) and Vol.2 (1969)' Department of Economics, RSSS, Australian National University)
A1.5. The purpose of this file was to construct 'pure' populations by by-passing censuses and the adjustments made from time to time to conform to census counts of the population. For example, it was possible to estimate the population at 30 June 1991 by commencing from the 1921 census count and, using the cohort component method, updating only for subsequent demographic events. Any census could be used as the base, and, once chosen, could be updated for any required number of years. In particular, though, these 'pure' populations were compiled for census dates to enable comparison with the census count at that date.
A1.6. For the years since 1976, the Databank has been maintained and updated by the ABS, and is now called the National Demographic Databank. It is a useful supplement to the PES as a measure of the quality of census data.
A1.7. Revisions have been made to the data base to conform to the changes, as from 1971, in the definition of 'population' ie. defining population as the number of people who usually reside in Australia (see Chapter 1). These revisions entailed the exclusion of all short-term overseas arrivals and departures back to 1925. (It was not possible to go back to 1921 because data for short-term movements were not available for years prior to 1925).
A1.8. Although it would have been possible to maintain the old series as well as the revised one, it was decided that only the revised series would be maintained and updated for future use.
A1.9. For analysing 1996 census undercount, an age range of 0 to 69 years could be derived for the population estimate from the Databank without using the 1921 Census as the base (since all such persons would have been born after 1921). This age range was considered sufficient for the purposes of assessing the 1996 census undercount, while the PES alone was available for the higher age groups. The problem of not having short-term migration data before 1925 is thus not applicable for these ages.