Household and Family Projections, Australia methodology

1 This publication contains 2016-based projections of households, families and living arrangements for the period 2016 to 2041. The projections are available for Australia, states and territories, capital cities and rest of states. These projections supersede the 2011-based series published in Household and Family Projections, Australia, 2011 to 2036 (cat. no. 3236.0) in March 2015.

2 Capital cities and rest of states are the GCCSA level of the Australian Statistical Geography Standard (ASGS) 2016 (cat. no. 1270.0.55.001). The ASGS defines the whole ACT as a capital city area, thus projections were not produced below the Territory level for the ACT. The projections for Australia include Christmas Island, Cocos (Keeling) Islands, Jervis Bay Territory and Norfolk Island, collectively referred to as Other Territories.

3 A household is defined as one or more persons, at least one of whom is at least 15 years of age, usually resident in the same private dwelling. Although in reality some living arrangements are fluid, the Census assigns each person to a single household. The projection method therefore treats households as discrete units.

4 These projections use a propensity method based on that developed by McDonald and Kippen and used in Household Trends and Projections: Victoria, 1986-2011. The method identifies propensities from the Census of Population and Housing for persons to be in different living arrangements. Trends observed in the propensities over the last four Censuses are assumed to continue into the future, and applied to a projected population. Numbers of families are then derived from the projected living arrangements of the population, and numbers of households are derived from the projected numbers of families. For details see Technical Note - 'Method'.

5 The projection results are not predictions or forecasts. They are illustrations of the change in the numbers of households and families that would occur if assumptions about future trends in living arrangements were to occur.　

6 The assumptions are based on past trends in living arrangements. No assessment has been made of changes in social and economic conditions which may influence future living arrangements. There is no certainty that any of the assumptions will be realised.

7 Sources of data for the projections are:

• The 2001, 2006, 2011 and 2016 Censuses of Population and Housing
• Estimated resident population at 30 June 2016 and 30 June 2017
• Estimated households at 30 June 2016
• ABS population projections at 30 June 2018 to 2041
   

8 Three Census items are used to obtain living arrangement propensities. These are household composition, family composition, and relationship in household of the usual resident population.

9 The Census only collects information on the relationship of each household member to person 1 and/or person 2 on the Census form. Relationships between other members of the household (e.g. between person 3 and person 4) are not captured. This may lead to an underestimation of the number of families, as well as relationships within the household and family type.

10 Person-level ‘relationship in household’ data within the Census is based on place of enumeration. This means persons temporarily absent from the household were excluded from the propensities due to limited relationship information. Visitors to households on Census night were also excluded from the propensities because they are not included in family coding in the Census.

11 Estimated Resident Population (ERP) published in Australian Demographic Statistics (cat. no. 3101.0) is used as the 2016 and 2017 population base (projected population is used for later years).

12 ERP is based on Census counts of usual residents, with adjustments made for Census undercount (people present in Australia on Census night but missed by the Census count) and usual residents of Australia who were temporarily overseas at the time of the Census.

13 ERP does not distinguish between people who live in private versus non-private dwellings (NPDs). Household and family 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 was identified and excluded from the calculation of numbers of households and families. People living in NPDs are included in the analysis of living arrangements.

14 Estimated households are used to constrain the base year (2016) living arrangement propensities calculated from the 2016 Census. For details see Technical Note - 'Method'.

15 The 2016 estimated households in Australia in this publication (9,204,635) is greater than the 2016 Census count of households (8,286,077) published in various 2016 Census publications. The difference is a combination of dwelling undercount and dwelling misclassification in the Census and persons that were temporarily overseas on Census night. For the differences between Census counts and ERP see Explanatory Notes 11 to 13.

16 Projected number of persons reflect the Series B population projections published in Population Projections, Australia, 2017 (base) - 2066 (cat. no. 3222.0)

17 Series B assumes ‘medium’ levels of future births, deaths and migration, based recent historical trends:

• the total fertility rate for Australia declining to 1.8 babies per woman in 2027, and remaining constant thereafter;
• life expectancy at birth will reach 82.1 years for males and 85.5 years for females by 2041;
• net overseas migration will reach 225,000 people per year by 2027 and remain constant thereafter;
• net interstate migration gains for some states and territories, and losses for others.

18 Only one population projection series is used. This means that the differences between the three series of household, family and living arrangement projections only reflect the different living arrangement assumptions, rather than differences in the underlying population.

19 Chapter 9 – ‘What If . . .’ discusses the results of using two other projections of the population, the Series A and Series C projections from Population Projections, Australia, 2017 (base) - 2066 (cat. no. 3222.0), as the assumptions about the future population of Australia and the states and territories.

21 Usually, in this publication, figures of less than one million are rounded in the text to the nearest hundred while figures of more than one million are rounded to the nearest hundred thousand.

22 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 2016 to 2041), in order to make differences between projected numbers (of households, families and people in different living arrangements) more discernible.

23 Some state and territory governments produce household projections, which may be a useful complement to the projections in this publication. These are typically based on different methods and are produced with specific attention to the observed and expected trends of that state, without the requirement of being constrained to Australia level data.

24 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.

25 As well as the statistics included in this and related publications, the ABS may have other relevant data available on request. Inquiries should be made to the National Information and Referral Service on 1300 135 070.

1 These projections use a propensity method based on that developed by McDonald and Kippen and used in Household Trends and Projections: Victoria, 1986-2011. The method identifies propensities from the Census of Population and Housing for persons to be in different living arrangements. Trends observed in the propensities over the last four Censuses are assumed to continue into the future, and applied to a projected population. Numbers of families are then derived from the projected living arrangements of the population, and numbers of households are derived from the projected numbers of families.

2 The method for obtaining household and family projections is outlined below. The method consists of five main steps:

• Step 1: Calculate living arrangement propensities from Census
• Step 2: Calculate assumed future living arrangement propensities based on past trends
• Step 3: Derive projected living arrangements from assumed future propensities
• Step 4: Derive projected numbers of families from the projected living arrangements
• Step 5: Derive projected numbers of households
   

3 2016 living arrangement propensities were calculated for each five-year age group (0–4, 5–9, …, 85 and over). The propensity is the Census counts of people (for each age group) in that living arrangement type, divided by the total Census count for each age group. There are 15 living arrangement types, as shown in summary tables in the page 'Types of Living Arrangements'.

4 Living arrangement propensities were calculated using a large proportion (91% in 2016) of the Census count of Australian usual residents. Categories of people included in the 2016 Census counts but excluded from the calculation of living arrangement propensities were:

• imputed records.
• people coded to 'other non-classifiable' relationship in household.
• visitors in the private dwelling in which they were enumerated, who were usually resident elsewhere in Australia.
• people enumerated in non-private dwellings who were usually resident elsewhere in Australia.
   

5 People included in the projected population but not included in the living arrangement calculation 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.

6 Distributing by this method is likely to lead to an underestimation of lone-person households. It is less likely that a lone-person household will be enumerated at home on Census night because there are fewer people in the household. This bias is addressed in part by replacing Census counts of lone-person households with lone-person household estimates in the propensity calculations.

7 Census-based living arrangement propensities in the base year (2016) are reconciled to align with the estimate of households published in Australian Demographic Statistics (cat. no. 3101.0) for both total households and lone-person households. These estimates are based on Census counts but have been adjusted to account for person and dwelling undercount, dwelling misclassification and residents temporarily overseas on Census night.

8 The difference between the total numbers of households and lone-person households was proportionally distributed to group and family 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 (paragraphs 17 to 18) in reverse. For example, to obtain the number of persons in a group household, the ‘reconciled’ number of group households was multiplied by the average size of group households in the 2016 Census.

9 The 2016 reconciled counts by living arrangement were proportionally distributed across the five-year age groups according to the 2016 living arrangement propensities (as described in paragraph 3). The distribution was constrained to sum to 30 June 2016 ERP (by age) and the reconciled totals for each living arrangement.

10 Reconciled base year living arrangement propensities are used in step 2 as the basis for the projected propensities.

11 A line of best fit was calculated for each living arrangement (for each age group), using the observed 2001, 2006, 2011 and 2016 propensities. If the fitted straight line produced an r-squared value of less than 0.2, it was assumed there had been no change in propensity between 2001 and 2016, and therefore the 2016 propensity for that living arrangement for that age group was held constant throughout the projection period in all three assumption series. The annual rate of change for each line of best fit was calculated.

12 Assumed rates of change were applied to the 2016 reconciled propensities to give projected propensities for the period 2017 to 2041 for each living arrangement (for each age group). This produced three series of projected propensities. Series I holds 2016 propensities constant over the whole period, series II assumed the observed trend would continue at a decreasing rate, and series III continues at the full rate of change over the whole projection period. For more information see the page 'Assumptions'.

13 Projected propensities were adjusted to ensure they sum to 100% in each five-year age group using iterative proportional fitting.

14 The projected propensities were then applied to the projected population of each age group to produce the projected number of people in each living arrangement, for each year from 2017 to 2041.　

15 Propensities were produced independently for Australia, the states, territories, capital cities and each balance of state. They were then applied to the relevant projected population.　

16 State-level data was constrained to sum to the Australian living arrangement totals and five-year age group totals. This was done using two-dimensional iterative proportional fitting. The process was then repeated to constrain capital city and rest of state data to state data. For a more detailed description of iterative proportional fitting, see Purcell, N. J. and Kish, L., (1979), Estimations for small domains, Biometrics, 35, pp. 365-384.

17 The number of couple families (with or without children) was calculated as half the number of partners in couple families (with or without children). The number of one-parent families was calculated as the number of male lone parents plus the number of female lone parents. The number of 'other families' was calculated by dividing the number of persons living in 'other families' by the average size of 'other families' at the 2016 Census. See paragraph 22 for the average size of ‘other families’ by part of state.

18 The 2016 estimate of the number of families in Australia in this publication (6,659,038) is greater than the 2016 Census count of families (6,070,316) published in various 2016 Census publications. The difference is a combination of net undercount in the Census and families that were temporarily overseas on Census night. For the differences between Census counts and ERP see Explanatory Notes 11 to 13.

19 Family households can contain more than one family. The number of family households was calculated as the number of families multiplied by the ratio in the 2016 Census of family households to families – see paragraph 22 for ratios by part of state.

20 The number of lone-person households was calculated as the number of male lone persons plus the number of female lone persons. The number of group households was calculated by dividing the number of persons in group households by the average size of group households at the 2016 Census. See paragraph 22 for average group household size by part of state.

21 Total projected households is the sum of all household types.

22 The ratios mentioned in paragraphs 17, 19 and 20 above are presented in the following table:

23 The propensities were derived using the area boundaries according to the relevant ABS Geography classification of each capital city and rest of state area as they existed at the relevant reference points (2001, 2006, 2011 and 2016). Previous analysis suggests 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.