|Page tools: Print Page Print All RSS Search this Product|
Limitations of census data for preparing projections
15 The census question on relationships in the household collects information on only one level of household relationship, that of each person in the household to Person 1 and/or Person 2 on the census form. In situations where household members are related to each other but not to Person 1 or Person 2, then that second level of relationships may be lost. This may lead to a distortion and underestimation of the number of families, as well as relationships within the household and family type.
16 Person level data within the census is based on place of enumeration. For 1991 and 1996, persons temporarily absent from the household on Census night were excluded from the propensities as they were only categorised by three broad age groups only, and therefore could not be allocated to household, family and relationship in household type, by five-year age groups. For consistency, persons temporarily absent from the household in the 2001 Census were also excluded from the propensities. Visitors to households on Census night were also excluded from the propensities because they are not included in family coding in the census.
Estimated resident population
17 The ABS publishes quarterly estimates of the ERP at the national and state level, and annually at the Statistical Local Area (SLA) level. The ERP is based on census counts of usual residents. Account is taken of census underenumeration and the number of Australian residents temporarily overseas at the time of the census.
18 ERP is obtained by adding to the estimated population at the beginning of each period the components of natural increase (on a usual residence basis) and net overseas migration. For states and territories, account is also taken of estimated interstate movements involving a change of usual residence. After each census, estimates for the preceding intercensal period are revised by incorporating an additional quarterly adjustment (intercensal discrepancy) to ensure that the total intercensal increase agrees with the difference between the estimated resident populations at the two respective census dates.
19 The ERP does not distinguish between persons resident in private and NPDs. Family and household projections are concerned exclusively with the population usually resident in private dwellings. Therefore, for the purposes of these projections, the proportion of the census count in NPDs is identified and excluded from the calculation of numbers of families and households but they are included in the analysis of living arrangements.
20 The ABS produces population projections every two to three years, with the most recent set being published in Population Projections, Australia, 2002 to 2101 (cat. no. 3222.0). Projections of the population by age and sex are produced for Australia, the states and territories and capital city/balance of state, excluding Australian Capital Territory balance of state.
21 The projections are based on a combination of assumptions on future levels of births, deaths and migration to arrive at the size, structure and distribution of Australia's population into the next century. Although 72 projection series are available, three main series are selected for analysis.
22 Using only one population projection series as the future population of Australia ensures that differences in the projections of households, families and living arrangements of people reflect changes in living arrangements only. The Series B population projection from Population Projections, Australia, 2002 to 2101 (cat. no. 3222.0) has been chosen as the future population of Australia and the states and territories, and assumes medium levels of fertility (with the total fertility rate for Australia declining to 1.6 babies per woman by 2005–06), declining improvements in life expectancy at birth, medium levels of net overseas migration (with annual net overseas migration gains of 100,000 people), and medium levels of net interstate migration.
23 Chapter 5—What If...? discusses the results of using two other projections of the population, the Series A and Series C projections from Population Projections, Australia, 2002 to 2101 (cat. no. 3222.0), as the assumption about the future population of Australia and the states and territories.
Overview and assumptions
24 There are a number of techniques which may be used for projecting household and family numbers, including both static and dynamic models. The ABS uses a propensity method which identifies propensities from the Census of Population and Housing for people in each five-year age group to be living in different living arrangement types. Trends observed in the propensities over the last four censuses are then projected forward and applied to the projected population. Projected numbers of households and families are derived from the projected living arrangements of the population.
25 The projection method initially produces household numbers in the base year that differ from ABS household estimates published in Australian Demographic Statistics (cat. no. 3101.0). As a consequence, before the census trends can be projected forward, living arrangements in the base year must be reconciled such that the number of households derived from them are consistent with ABS household estimates. To ensure the discrepancy that arises between the two methods is not carried through the projected years, reconciled base year living arrangement propensities, as opposed to census propensities, are used as the basis for obtaining projected propensities which are then applied to the projected population to derive projected living arrangements.
26 While the use of reconciled living arrangement propensities results in the number of households from the projection method matching published ABS household estimates for 2001, the population of people living in private dwellings, derived by subtracting the number of people living in NPDs from the total population, differs slightly from the 2001 estimated resident population in households data published in Australian Demographic Statistics (cat. no. 3101.0). This is due to different methods for calculating the population in NPDs. In this publication the number of people living in NPDs was calculated directly from 2001 Census propensities, similar to the other living arrangements, whereas in Australian Demographic Statistics further adjustments were made to the number of people living in NPDs in the 2001 Census, to include people sleeping in tents, sleeping out, and in improvised homes where the collection method also suggested a person was not living in a private dwelling.
27 A household is defined as a group of two or more related or unrelated people who usually reside in the same dwelling, who regard themselves as a household and who make common provision for food or other essentials for living; or a person living in a dwelling who makes provision for his or her own food and other essentials for living, without combining with any other person. The census categorises households as discrete units, thus the projection method is based on the assumption that households are discrete units. It is acknowledged that some living arrangements are fluid and that because of study, work or shared care arrangements a family member may be associated with more than one household. For example, full-time students living away from home may still be dependent on their family household.
28 This method assumes that trends observed over the four censuses accurately reflect underlying changes in living arrangement characteristics over that time period.
29 Living arrangement propensities were calculated using a large proportion (95% in 2001) of the census count of Australian usual residents. People not counted in the census, or not included in the classification of living arrangements, but included in the projected population were proportionally distributed across the living arrangement types. This assumes that these people have the same living arrangement distribution as those included in the calculation of the propensities.
30 Distributing by this method is likely to lead to an underestimation of lone person households. If an entire household is absent on Census night, that household can be missed. However, if some members of the household are at home, the household type can be coded based on the information of those members remaining at home, and the information provided for members temporarily absent. It is therefore more likely for households to be missed if the household consists of only one member.
31 The method for obtaining household and family projections is outlined in detail in paragraphs 32 to 52 below. The procedure consists of four main parts; (1) obtaining the rate of change in living arrangement propensities; (2) reconciling base year living arrangements to align with ABS household estimates; (3) obtaining projections of living arrangements; and (4) projecting the number of families and households.
Step 1: Using census data to obtain the rate of change in living arrangement propensities
32 Census counts from the past four censuses, classified by living arrangement type and five-year age group, were used to obtain the rate of change in living arrangement propensities. Living arrangements included 15 categories as shown in table 6.4.
Step 1.1: Population distribution by living arrangement
33 Using the 1986, 1991, 1996 and 2001 Census datasets, census counts were classified by living arrangement and five-year age groups. Visitors to households, overseas visitors and people in non-classifiable households were excluded, as well as imputed or system created records for 2001 as they contain no household or family information.
Step 1.2: Calculation of propensities
34 From the distribution produced at step 1.1, the propensity of persons in each five-year age group to live in each living arrangement was calculated using the total count of persons by five-year age group as the denominator. Propensities for the 1986, 1991, 1996 and 2001 Censuses are shown in Appendix 1.
Step 1.3: Rate of change in propensities
35 For the purpose of trend analysis, the 1986 and 1991 living arrangement counts were recoded to achieve consistency with the 1996 and 2001 definitions (see paragraphs 9–14).
36 A line of best fit was calculated for each living arrangement and age group, using the observed 1986, 1991, 1996 and 2001 propensities (calculated in step 1.2). The annual rate of change for each line of best fit was calculated. A condition to measure the closeness of fit of each straight line was employed. If the fitted straight line produced an r-square value of less than 0.2 it was assumed there had been no change in propensity between 1986 and 2001, and therefore that the 2001 propensity for that living arrangement and age group would hold throughout the projection period.
Step 2: Reconciling base year living arrangements with ABS household estimates
37 The number of persons by living arrangement type in the base year (2001) were adjusted to ensure that the number of households derived from them were consistent with ABS 2001 household estimates.
Step 2.1: Obtaining the number of households from the projection method
38 The 2001 census propensities were applied to 2001 ERP to obtain 2001 ERP by living arrangements and five-year age groups. From these living arrangements the number of households and families were obtained using the method outlined in step 4.
Step 2.2: Reconciling living arrangements with household estimates
39 The number of lone person households and total households obtained from the projection method were replaced with 2001 ABS household estimates. The difference between the total number of households and lone person households was proportionally distributed amongst family and groups households. The number of families and then persons by living arrangement were derived from the scaled household estimates, using the method outlined in step 4 in reverse. For example, to obtain the number of persons in a group household from the number of group households, the reconciled number of group households was multiplied by the average size of a group household based on the 2001 census (as opposed to dividing the number of persons in group households by this ratio to obtain the number of group households).
40 The 2001 reconciled counts by living arrangement were proportionally distributed across the five-year age groups according to the 2001 census propensities (as generated in step 1.2). The distribution was restricted by 2001 ERP (by age) and the reconciled totals for each living arrangement.
Step 2.3: Calculating reconciled propensities
41 The propensity to live in each living arrangement and age group from the reconciled 2001 counts by living arrangement and age group was calculated. Reconciled propensities are shown in Appendix 2.
Step 3: Obtaining projected number of persons by living arrangement
42 The 2001 reconciled propensities were used as the basis for obtaining the projected propensities.
Step 3.1: Applying census trends to obtain projected propensities
43 The rates of change generated in step 1.3 were applied at the fractions specified in the assumptions (zero, reducing from one, and one, for Series I, II and III respectively), to the 2001 reconciled propensities (generated in step 2.3). This produced three series of 'projected propensities'. Projected propensities were subsequently adjusted to add to 100% in each five year age group.
Step 3.2: Applying projected propensities to the projected population
44 For each living arrangement and age group, the projected propensities were then multiplied by the projected population of each age group to determine the three series of projected population by living arrangement type.
Step 3.3: Constraining of sub-state and state totals to Australian totals
45 Propensities were produced for Australia, each state and territory, each capital city and each balance of state (excluding ACT balance of state) and applied to the respective projected population. As the propensities were projected independently at each geographical level, after they were applied to the projected population their sum did not equal the distribution of living arrangements for Australia. In order to ensure consistency in the projected number of persons, constraints were applied to person level data to ensure that the sum of the sub-states and states add to that of Australia.
46 Constraining was conducted using two-dimensional iterative proportional fitting (IPF). To constrain to Australian totals, a matrix was constructed for person counts for each five-year age group for each state/territory and living arrangement type. The figures in the body of the matrix were proportionally scaled across rows and columns simultaneously constraining to the Australian living arrangement totals in the final column and age totals in the final row. The scaling process was repeated several times until stability was attained within the matrix. Decimals produced by the scaling process were then rounded, with column and row totals preserved. The final matrix was fed back into the extrapolation process, forming the base for the next year's projections.
47 This procedure was then conducted at the sub-state level. The same process was followed, with the matrix constrained simultaneously to state or territory living arrangement totals and sub-state age totals. For a more detailed description of the IPF procedure, see Purcell, N J and Kish, L (1979) Estimations for small domains, Biometrics, 35, pp. 365–384.
Step 4: Calculating the number of families and households
48 The projected number of households and families were derived from the three series of living arrangement projections. When calculating the number of 'other families', family households and group households, ratios from the 2001 Census were applied to account for varying household and family sizes in these categories. These ratios were generated and applied at the capital city/balance of state level, from which state and Australia level data were aggregated. The ratios are presented in the following table.
2001 CENSUS RATIOS
Step 4.1: Deriving numbers of families from the living arrangement types
49 The number of couple families (with or without children) were calculated as half the number of partners in couple families (with or without children). The number of one-parent families were the number of male lone parents plus the number of female lone parents. The number of 'other families' were calculated by dividing the number of related individuals in other families by the average size of this family type.
Step 4.2: Converting families to family households
50 Family households can contain more than one family. In order to produce numbers of households, families were converted to households using the family households to families ratio.
Step 4.3: Deriving numbers of non-family households
51 Numbers of non-family households were also derived from the living arrangement types. The number of lone person households were calculated as the sum of the number of male lone persons plus the number of female lone persons. The number of group households were calculated by dividing the number of persons in group households by the average size of group households at the 2001 Census.
Step 4.4: Deriving the total number of households
52 Family and non-family households were added to produce the total number of households.
DIFFERENCE BETWEEN 2001 ESTIMATES OF FAMILIES IN THIS PUBLICATION AND 2001 CENSUS COUNTS OF FAMILIES
53 It should be noted that 2001 estimates of families in Australia and the states and territories in this publication differ from the number of families according to the 2001 Census of Population and Housing. The 2001 estimate of the number of families in Australia in this publication (5,345,971) is greater than the 2001 Census count of families (4,936,828) published in Census of Population and Housing: Selected Social and Housing Characteristics, Australia (cat. no. 2015.0) and other 2001 Census publications. This is due to estimates of families in this publication being based on 2001 estimated resident population as well as 2001 Census living arrangement propensities. The 2001 ERP of Australia is greater than the 2001 Census count of people in Australia as it includes further estimates of net Census undercount and residents temporarily overseas, and as a result, the number of families based on 2001 ERP is also greater than the number of families from the 2001 Census.
54 The propensities were derived using the area boundaries as they existed at the relevant reference points (1986, 1991, 1996 and 2001). It is considered that the use of such propensities (rather than based on revised areas) would not have had a significant effect on the comparability of the propensities over time.
55 Generally, in this publication figures of less than one million are rounded in the text to the nearest thousand while figures of more than one million are rounded to the nearest one hundred thousand. Figures in tables in Chapter 6 have been rounded to the nearest hundred.
56 Changes in population over time are commonly discussed in terms of average annual growth rates. In this publication, however, changes are for the most part presented as percentage increases or decreases over the entire 25 year period (from 2001 to 2026), in order to make differences between projected numbers (of households, families and people in different living arrangements) more discernible.
COMPARISON WITH STATE GOVERNMENT HOUSEHOLD PROJECTIONS
57 The following table has been provided for comparative purposes. It gives details of ABS and state and territory government household projections at the capital city and balance of state level for 2011. Note that ABS and ACT government projections are 2001-based, Victorian and Western Australian projections are 1996-based, and New South Wales and Queensland projections are 1991-based. Not all state and territory governments produce household projections.
PROJECTED NUMBER OF HOUSEHOLDS
58 ABS publications draw extensively on information provided freely by individuals, businesses, governments and other organisations. Their continued cooperation is very much appreciated; without it, the wide range of statistics published by the ABS would not be available. Information received by the ABS is treated in strict confidence as required by the Census and Statistics Act 1905.
59 Users may also wish to refer to the following ABS publications:
Australian Demographic Statistics (cat. no. 3101.0) — issued quarterly
Census of Population and Housing: Selected Social and Housing Characteristics, Australia (cat. no. 2015.0)
Demographic Estimates and Projections: Concepts, Sources and Methods (cat. no. 3228.0) included in Statistical Concepts Library available on the ABS web site.
Household and Family Projections, Australia, 1996 to 2021 (cat. no. 3236.0)
Population Projections, Australia, 2002 to 2101 (cat. no. 3222.0)
60 Current publications and other products released by the ABS are listed in the Catalogue of Publications and Products (cat. no. 1101.0). The Catalogue is available from any ABS office or the ABS web site <https://www.abs.gov.au>. The ABS also issues a daily Release Advice on the web site which details products to be released in the week ahead.
61 As well as the statistics included in this and related publications, additional information is available from the ABS web site <https://www.abs.gov.au> by selecting Themes then Demography.
These documents will be presented in a new window.