• No results found

Attributes of the Population – Census and Related Data

Chapter 3: Data Sources

3.3 Attributes of the Population – Census and Related Data

3.3.1 ABS Population Censuses 1996 and 2001

Data on population age, gender, education, income, unemployment, place of birth, Indigenous Australians, and the area of each SLA were drawn from the 1996 and 2001 population censuses (Australian Bureau of Statistics 1997; Australian Bureau of Statistics 2002).

Population attributes were summarised as the percentage of the population in the relevant categories, except for income where mean personal income was used.31 The population basis for all years was the estimated resident population (Section 3.3.3). Ages again were grouped, in this case to those people aged less than 15 years, those aged 15 to 64 years and those aged 65 years and over.

The education measure is the proportion of the population with a post-school diploma or higher qualification, which is a measure of those with post school qualifications. Place of birth was classified as ‘Australian born’ and ‘other’.

Information was extracted on the proportion of the employed population engaged in white collar versus blue collar industries.32 This was an environmental factor and likely to influence the attractiveness of an area to GPs as a place to work, although it may also have had some influence on health, with blue collar workers more inclined to face workplace injury. The Australian Bureau of Statistics (2006d) shows that workers in manual industries such as agriculture, forestry and fishing and mining have materially higher rates of work-related illness and injury (90–100 cases per year per 1000

employed persons, compared to around 20 cases per 1000 employed persons in the finance industry).

Percentages in the relevant categories and mean personal income were linearly interpolated at the SLA level for years between 1996 and 2001, and linearly

extrapolated for 2002 and 2003. Data for higher levels of aggregation were obtained by

31 Income is price adjusted using the CPI to reflect purchasing power (Australian Bureau of Statistics

adding interpolated and extrapolated numbers from the SLA level prior to calculating ratios.

3.3.2 Socio-economic Classifications

The ABS develops a range of measures of socio-economic wellbeing from every census (Australian Bureau of Statistics 2003a). The structure of the indexes changed between 1996 and 2001, and the only general index available in both years was the Index of Disadvantage, which is used in this study. This index includes all available census variables that either reflect or measure disadvantage. Although it is formally titled the ‘Index of Disadvantage’, throughout this thesis the index is referred to as the SEIFA (the Socio-Economic Index for Areas), which is the more generic term applied to these indexes.

The SEIFA was designed to have a mean of 1,000 and a standard deviation of 100 across all Census Collection Districts (CDs), so it had approximately the same mean and a smaller standard deviation across SLAs which were composites of CDs. To obtain data for non-census years, the original SEIFA scores have been interpolated and extrapolated as for the other census variables.

As the SEIFA is derived by ABS using a principal components methodology, it provides an ordering of the SLAs but is not an interval variable. “Ordinal measures allow the CDs to be ranked but distances between two CD values (with equal differences in SEIFA scores) are not necessarily equivalent” (Australian Bureau of Statistics 2003a, p. 71). In other words, a difference of 100 points in the index value did not always mean the same difference in the level of disadvantage. It was therefore not appropriate to use the raw SEIFA scores directly in the modelling. The SEIFA scores have been classified into quintiles (distributed so that equal proportions of the total population were in each category), and included as a series of five indicator variables reflecting the quintile to which the area belongs.

When it was necessary to aggregate SEIFA scores to broader geographic levels, the original SEIFA scores were combined, weighted by the population of the area, before quintiles were derived.

3.3.3 Population Count Data

Population data for each SLA for each year were extracted from the ABS estimated resident population data (Australian Bureau of Statistics 2003d; Australian Bureau of Statistics 2004).33 This estimated resident population was used as the denominator in estimating percentages of aged, young, etc., and in estimating the population density.

3.3.4 Levels of Remoteness

Levels of remoteness were defined as part of the ABS Australian Standard Geographic Classification (Australian Bureau of Statistics 2003b; Glover, Hetzel et al. 2006). The

Accessibility Remoteness Index of Australia (ARIA) allows the comparison of information about regions based on their access by road to service centres (towns) of various sizes. While the index is a continuous variable suited to some forms of research, for the purposes of this study discrete categories were more appropriate.

The ABS grouped the index into five categories: Major Cities, Inner Regional, Outer Regional, Remote and Very Remote. Each SLA was allocated to a remoteness category based on 2001 data.34 While some SLAs would have changed categories between 1996 and 2003 (especially on the metropolitan fringes), the regional classification based on 2001 data has been used in all analyses (hence the level of remoteness could not be used in panel studies).

3.3.5 Numbers of People Working in SLAs

Many patients obtain their GP services (or at least some of their GP services) near their place of work rather than their place of residence. This applies in particular for people who work in the central business districts (CBDs) of major cities.

33 Concordances were required to convert some data to 2001 SLAs. 34

Data privately provided by AIHW showed the proportion of each 2001 SLA in the various remoteness categories based on CD populations. Each SLA was allocated to the region in which the majority of its

Data were extracted from the 2001 Population Census on the number of people employed in each SLA. This was divided by the population of the area to give an ‘employment rate’, which was substantially more than one in a number of cases (e.g. CBDs) while in other areas (e.g. dormitory suburbs) it was substantially below one.

Data on the number of workers were available for only a small number of selected areas for 1996, so it was not possible to use this data in panel estimation. However data from the 2006 Census will become available in due course, and will permit use of this

variable in future panel studies.