A New Measure of Average Household Size | Bulletin – March 2023 (2024)

Introduction

Average household size (AHS) – the average number of adults and children living in a home – isa key determinant of underlying demand for housing. For example, a decline in AHS means more householdsare being formed and there is therefore greater demand for housing for a given level of populationgrowth. Changes in AHS can by driven by structural factors, such as shifts in demographics and householdpreferences for how much space people want; changes in AHS can also occur in response to cyclicalconditions, such as changes in housing prices and rents (Kohler and van der Merwe 2015). During the COVID-19 pandemic, changes in AHS were an important margin of adjustment forthe housing market (Ellis 2022).

Existing AHS measures are published infrequently and with a long lag, which does not allow for a timelyassessment of the interaction between changes in AHS and housing market conditions. This articleintroduces a timelier measure of AHS – developed by the authors – that can be calculated fordifferent groups in the Australian population, including by geographic area and householdcharacteristics.

A new measure of AHS using Labour Force Survey data

Background

Research on household size and household formation in Australia has tended to use data from the Census ofPopulation and Housing, which is conducted by the Australian Bureau of Statistics (ABS) every five years.While the Census is an invaluable source of data on longer run trends in household formation, it is notsuitable for monitoring recent and higher frequency changes in household formation such as those thatoccurred since the onset of the pandemic. Other sources of household-level data – including, theSurvey of Income and Housing (SIH) and the Household, Income and Labour Dynamics in Australia (HILDA)survey – are also not available frequently and timely enough for real-time monitoring of householdformation.

Recently, the Bank has started constructing and monitoring a new series on AHS (Graph1). This newmeasure uses the data underlying the ABS’s monthly Labour Force Survey (LFS). While the LFS was notspecifically designed to measure AHS, it aligns closely with and addresses many of the shortcomings ofthe existing measures.^[1] Hereafter, the new measure constructed by the authorsis referred to as the ‘LFS Measure’.

The LFS Measure is timely: it is available on a monthly basis and can be updated six weeks after the endof the period that each survey references (e.g. if the survey is conducted in the first half of March,estimates for March are generally available by the fourth week of April). As a result, the LFS providesthe most up-to-date estimate of AHS, with the latest data suggesting that the AHS remained aroundhistorically low levels at 2.49people per household in January 2023. The LFS Measure also has theadvantage of being based on a large nationally representative sample, with monthly data underpinning themeasure available back to July 1983. The long history allows us to adjust for regular seasonal patternsand to understand trends and cycles over time. Table1 summarises the timeliness of the variousmeasures of AHS. AppendixA provides details on the various measures and their differences.

Methodology

The LFS Measure is simple to construct. As part of the LFS, the ABS collects information on thecharacteristics of each household, including the number of adults and children who usually live in thedwelling. These questions have been asked every month since mid-1983. Both the LFS and Census estimatesof AHS in Graph1 count the number of persons who are usually resident in the same private dwellingand exclude visitors and people who usually live in non-private dwellings (such as nursing homes, hotelsand boarding schools).

The LFS Measure is calculated by the authors using the following methodology:

1. The number of residents in each household is counted by summing the number of adults and childrenwithin each occupied private dwelling for each month of the survey. Visitors to the household areexcluded from this count.
2. The average number of people in each household across Australia is calculated for each month. Inorder to ensure this is broadly representative of the whole population, our average is weighted bythe probability of the adults appearing in the LFS.
3. The series is seasonally adjusted to minimise the effects of any variations arounds holidays, theuniversity calendar or other seasonal events, although the seasonal effects are small.^[2]

The series can be revised over time if the ABS make changes to the population measures the LFS dataare aligned to.

An example of the code is provided in AppendixB, and graph data are available on request.

Recent developments in household size

At the onset of the pandemic

The LFS Measure shows that AHS picked up noticeably at the beginning of the COVID-19 pandemic (Graph2, top panel). Household sizes grew, on average,by around 2percent between February and September 2020, during which time a large share ofthe population was under lockdown.

The spike in AHS reflected compositional changes in households (Ellis 2022). The share of households withone person fell by 2percentage points, while the share of households with three or four peopleincreased by a similar amount (Graph2, bottom panel). In part, these shifts likely reflected alarge number of young adults moving home with their families, with the share of those aged 18–30years old living with their parents rising by5percentage points to a historical high (Graph3). These trends were likely driven by effortsto lower housing expenses at a time of labour market weakness and economic uncertainty, the shift toonline learning and work, as well as a desire to live with others amid social distancing outsidehouseholds.

From late 2020

The increase in AHS at the onset of the pandemic abruptly reversed in late 2020, with AHS declining overthe following two years from 2.55individuals per household to a historical low of2.48individuals by August 2022. This trend could reflect an increase in the demand for space, aspeople spent more time at home, both due to pandemic-related health restrictions and the increasedability for some to work from home.^[3] Population growth was also much lower than expectedduring the pandemic, due to lower net overseas migration and the closure of international borders. Thiscontributed to a period of high rental vacancy rates and lower advertised rents, particularly in Sydneyand Melbourne, in 2020. Greater affordability, combined with changed preferences and strong incomegrowth, likely encouraged individuals to form smaller households with more space per person over 2021 and2022.

The decline in AHS contributed to a sizeable increase in demand for the number of homes in Australia,which helped offset the relatively slow growth in the population during the pandemic (Graph4).^[4] A roughcalculation suggests that across the Australian population of more than 25million people, a declinein AHS of the magnitude observed between early 2020 and September 2022 (around 1percent,without any change in population growth) would alone imply an increase of around 120,000households.

The outlook for AHS is uncertain. AHS has increased slightly since the end of 2022, possibly in responseto tightness in many rental markets, particularly regional areas (see discussion below). But manystructural factors that have contributed to the formation of smaller households over recent decades– such as an aging population, falling fertility and marriage rates, and higher household incomes– are still relevant today. Furthermore, it is not clear whether the preference for additionalspace will endure.

Trends by location

AHS varies considerably by location. Capital cities, on average, have larger households than regionalAustralia, likely reflecting higher housing costs (Graph5). It may also reflect compositionalfactors. New households, such as those created by permanent arrivals from overseas, tend to be larger andmore likely to settle in capital cities than regional areas due to the proximity to employment andeducation opportunities.

The up-down cycle in AHS since the onset of the pandemic has largely been driven by developments incapital cities, in particular Sydney and Melbourne (Graph6). These cities experienced a sharpincrease in the AHS at the start of the pandemic. They had longer and more stringent lockdowns comparedwith other parts of Australia, which may have led to a greater shift and persistence in preferences foradditional space. Both cities were also more exposed to changes in overseas migration associated with theinternational border closure than other parts of the country. In part reflecting the decline in arrivals,Sydney and Melbourne experienced weaker rental markets at the onset of the pandemic relative to othercapital cities and regional areas, with large increases in rental vacancy rates and declines inadvertised rents by mid-2020 (Graph7).

See Also

Calculating Your Waist Circumference

AHS in Sydney and Melbourne declined over 2022 and has remained around historical low levels, alongside aconsiderable tightening in rental market conditions in both cities (Agarwal, Gao and Garner 2023). Bycontrast, there are some signs of AHS increasing or stabilising in some other capital cities, such asBrisbane, and regional areas, where rental markets have been tight for a longer period of time.

Conclusion

The new LFS Measure of AHS has the advantage of being timely, frequent and based on a large representativesurvey. Given the detailed demographic and employment-related information available in the LFS microdata,there are many other subgroups for which we could calculate AHS. This article has focused on changes inAHS following the onset of the pandemic and examined developments in living arrangements and in differentparts of Australia. However, the LFS Measure can also be used to examine how AHS has changed acrossdifferent types of households, including by industry and occupation of employment.

Other researchers can make use of the data and code developed as part of this article.^[5] The Bankwill continue to closely monitor developments in AHS using this new measure, as well as assessing itsrole as an indicator of underlying housing demand.

Appendix A: Various measures of AHS

The LFS Measure of AHS aligns with other measures (GraphA1). AHS steadily declined from around2.9people per household in the mid-1980s to be close to 2.6people by the late 1990s,consistent with Census data. AHS was relatively flat from the early 2000s until the mid-2010s, beforecontinuing on a downward trend over the 2020s, which is evident in all other AHS measures.

The consistent results reflects that these measures are conceptually similar, although there are a fewdifferences in scope and coverage. All estimates of AHS measure the number of persons who are usuallyresident in the same private dwelling. Most people who usually live in the household but were temporarilyabsent during Census night or the survey (e.g. for work, school, other purposes or because they weretemporarily overseas) are treated as being part of the household. These measures also exclude visitors tothe household. Furthermore, visitor-only households and people usually living in non-private dwellings(such as nursing homes, hotels, boarding schools and hospitals) are excluded from the average.

There are also small level differences between the LFS estimate of AHS and others that likely relate tovariation in scope and coverage of each source of data. For example, the SIH and the LFS both excludehouseholds in very remote geographical areas or with members of the Australian defence force, while thesehouseholds are included in the Census. However, this has only a small impact on the aggregate data, ashouseholds in very remote areas constitute a small portion of the population and households with defencepersonnel have mostly similar household sizes.

Appendix B: Sample code for Stata

A list of variable names and definitions can be found at ABS (2023).

********************************************************* Set up and cleaning********************************************************* Use LLFS microdata. The date has been recoded to a monthly variable. ‘hhid’,‘id’, ‘weight’ ‘date’ are the variable names for household id, individual id, representative weight and date.use lffsdata.dtakeep if date >= tm(1983m7)* Recode missing values for number of children aged 14 years or younger (nkid14h)replace nkid14h =. if nkid14h <0 & nkid14h !=.* Drop all visitors and non-private dwellingsdrop if urstatus!=1********************************************************* Generate the national LFS Measure of AHS********************************************************* Count the number of usual adult residents in the household. This excludes visitor from the count and assigns this figure to each person in the household.bysort hhid date: egen ahs = count(id)* Add the number of resident children to this figure, providing the total number of persons who are usually resident in the same private dwelling.replace ahs = ahs + nkid14h* Create a household-level weight, based on the average probability of each adult appearing in the LFS.collapse (mean) ahs weight, by(hhid date)* Calculate the average number of people in each household across Australia for each month.collapse (mean) ahs [weight = weight], by(date)