RDP 2014-03: Household Saving in Australia 2. Data

The 2003/04 and 2009/10 Household Expenditure Surveys are cross-sectional surveys of a nationally representative sample of households in Australia during the survey period.^[2] For each household, the surveys collect information on income and consumption, as well as a range of socio-demographic characteristics. These socio-demographic characteristics allow us to assess the saving behaviour of particular groups of households, which is not possible with aggregate data.

The ABS also conducted expenditure surveys in 1975/76, 1984, 1988/89, 1993/94 and 1998/99. We do not use these earlier surveys in our analysis since: (i) methodological changes render surveys conducted before 1998/99 less comparable to those from 1998/99 on; and (ii) the surveys conducted before 2003/04 omit important variables, such as household wealth, which can play a large role in influencing saving behaviour.

2.1 Definition of Income, Consumption and Saving

The HES collects detailed expenditure data using two methods: the ‘diary’ method and the ‘recall’ method. The diary method involves each household recording their regular expenditures in a diary over a two-week period. The recall method is used for expenditure on durable items; households are asked to remember how much they spent on these items over a given period. The surveys also contain data on personal and household income, which are collected through interviews. The most important quantitative data that we use are household income, consumption and saving.

Disposable income includes: wage and salary payments; tips; other labour income; farm income; income of unincorporated enterprises; net rental income; imputed rent for owner-occupiers^[3]; interest on savings; dividends; transfer income from the government, private institutions and other households; superannuation contributions by employers on behalf of employees; superannuation drawdowns by self-funded retirees; inheritance; gifts and other income from family members. Income is after tax and interest payments.

Note that the national accounts definition of income includes a number of items that are unavailable in the HES, the largest of which are imputed interest and current transfers to non-profit institutions serving households. We also cannot separately identify (and therefore exclude) capital draw-downs from investment earnings for self-funded retirees in the HES, so that income for self-funded retirees is overstated.

Consumption includes total expenditure on goods and services as well as imputed rent for owner-occupiers. Principal and interest repayments on debt, home capital improvement expenditure and life insurance and superannuation related expenses are not included in consumption.

Saving is calculated as the difference between disposable income and consumption. The main difference between our definition of saving and that from the national accounts stems from the different definition of income, as noted above. Note that our definition of saving captures only active saving and does not include any capital gains or losses.

2.2 Comparison of Aggregate and Micro Data

In order to use the household surveys to analyse the drivers behind the increase in the aggregate saving ratio, the surveys must be comparable with each other and with data from the national accounts. There were no major methodological changes between the 2003/04 and 2009/10 Surveys, so the two surveys should be comparable, and while the surveys do not capture all household consumption or income when compared with national accounts data, they capture a similar proportion of each. This implies that the aggregate saving ratio from the HES datasets should be consistent with the aggregate saving ratio from the national accounts.

Figure 2 compares the aggregate (mean) gross household saving ratio implied by the surveys with that implied by the national accounts, where income and consumption in the national accounts are defined so as to match the HES definitions where possible. For comparison we include the saving ratio from the 1998/99 Survey. Both measures show the saving ratio to be little changed between 1998/99 and 2003/04, then increasing sharply between 2003/04 and 2009/10.

2.3 Descriptive Analysis

Looking at the distribution of the saving ratio across households, the median saving ratio in 2003/04 was 5 per cent, while in 2009/10 it was 9 per cent. The shift up in the saving ratio evident across most of the distribution is consistent with the rise in the aggregate saving ratio over this period (Figure 3). The distribution of the saving ratio displays a long tail of negative saving ratios (negative skew). This is unsurprising as consumption is always positive, but income, which is the denominator of the saving ratio, can sometimes be close to zero, which leads to large negative saving ratios for some households.

Figure 4 shows how income, consumption and saving vary by household age in the 2003/04 and 2009/10 Surveys. Household consumption tends to track income closely, with both varying significantly over the life cycle, suggesting that households do not fully smooth their consumption, although Attanasio (1999) points out that the hump-shaped consumption profile is less pronounced after controlling for family size and composition. Between the 2003/04 and 2009/10 Surveys, saving increased especially for younger and older households, with income rising more than consumption for these groups.

Changes in household saving behaviour do not appear to be specific to certain levels of household wealth, with the saving ratio increasing across all wealth quintiles between 2003/04 and 2009/10 (Figure 5). Most (age-matched) income quintiles also saw a rise in saving between 2003/04 and 2009/10, with only the lowest income group recording a fall in saving (Figure 6).^[4]

This simple descriptive analysis suggests that relatively young and old households, but not middle-aged households, considerably increased their saving between 2003/04 and 2009/10, while a change in saving behaviour was evident across most wealth and income groups. While this could in part reflect the ageing of the population, we find that this ageing effect is not large enough to explain the large increase in the saving ratio over the 2000s. (Appendix A describes a simple decomposition model based on age and birth cohorts, and shows that the increase in saving cannot be attributed to these factors).^[5] Given this, we need to consider other explanations for the rise in the saving ratio.

Footnotes

The 2003/04 HES surveyed around 7,000 households, while the 2009/10 HES surveyed around 10,000 households. The sample used excludes those who give zero or negative values for income, and households where the household head is aged over 75 years. We also trim the top and bottom 2 per cent of the sample based on the saving ratio distribution. [2]

Imputed rent for owner-occupiers is determined using the methodology outlined in ABS (2008) for 2003/04; imputed rent using this methodology is already included in the 2009/10 HES. [3]

Age-matching involves splitting the households in each age group into separate income quintiles. The corresponding income quintiles from each age group are then combined, so that, for example, the lowest age-matched income quintile consists of those households that make up the lowest income quintile within each age group. Income quintiles are age-matched in order to separate age and income effects; for example, since post-retirement households are typically in the lower income quintiles, the saving behaviour of older households will govern the saving behaviour of the lower (non age-matched) income quintiles. [4]

This is not surprising given that most of the rise in the saving ratio occurred over a relatively short period, whereas the ageing of the population is a slow-moving process. Chamon and Prasad (2010) find a similar result in their study, while Browning and Lusardi (1996) argue that ageing is too slow to provide a sufficient explanation for the large decline in the US aggregate household saving ratio. [5]