RDP 9511: Superannuation and Saving 4. The Australian Experience

As in several other countries, the aggregate net household saving ratio in Australia, as conventionally measured, has steadily declined over the past two decades. Saving rose sharply in the early 1970s, before peaking at over 14 per cent of household disposable income in 1974/75. By the mid 1990s, the saving rate had fallen below 5 per cent.

It is widely believed, however, that part of the large rise in measured saving in the 1970s was purely mechanical – the result of an upward measurement bias that occurs during periods of high inflation. National accounts measures of saving do not adjust interest payments to households for that component which simply compensates savers for the decline in the real value of assets generating the income flows. As a result, net interest receipts (and hence saving) are higher during periods of high inflation than would otherwise be the case. A crude adjustment to the net saving rate is shown in Figure 1.

Even with the adjustment, however, the saving rate has still generally fallen over the past couple of decades. The fall, however, has not been uniform across all forms of saving. The superannuation component of measured saving averaged about 4 per cent of disposable income during the 1960s and 1970s, before rising to close to 8 per cent during the 1980s. More recently, superannuation saving declined to less than 3 per cent of disposable income. By implication, the aggregate of other forms of household saving fell substantially in the 1980s before recovering somewhat in the early 1990s.

In itself the observation that non-superannuation saving fell during a period when superannuation saving rose sharply does not necessarily imply a causal relationship. While changes in superannuation saving have the potential to explain part of the shifts in other saving, it may be that other factors, quite independent of superannuation, have caused much of the observed behaviour of other saving. The extended life-cycle/permanent income model outlined in Section 3 identifies a number of possible factors which, in addition to shifts in superannuation saving, may help explain patterns of saving in Australia over the past few decades.

4.1 Non-Human Wealth

In Australia, as in many other countries, non-human wealth increased sharply in the 1980s, after moderate growth in the preceding two decades.^[19]

Much of the rise, however, was a result of increases in the value of housing and it is unclear how important wealth effects have been in this case. Life-cycle theory predicts that higher housing wealth would decrease the propensity of home owners (usually older households) to save out of current income. However, housing is a relatively illiquid form of asset holding and it is likely that, in the past, borrowing restrictions prevented many households from optimally borrowing against housing wealth for consumption purposes. Financial deregulation may have made it easier for households to access dwelling wealth in recent years, but it is still unlikely that households would have responded to increases in dwelling wealth to the same extent that they would have if the increase in wealth had come from a rise in the value of more liquid financial assets.

It is also possible that the negative effect on saving by older households may have been partially offset by increased saving by younger households, because non-home-owners have to accumulate a substantial deposit before lending finance is available. The effect of higher house prices in the 1980s may have been to increase the amount of funds that younger households needed to accumulate as a deposit to purchase the more expensive housing (Lattimore 1994).^[20]

4.2 Human Wealth

Life-cycle/permanent income models also give an important role to human wealth effects but, without a clear understanding of how expectations of future labour income are formed, human wealth is difficult to measure.

The evolution of actual labour income, however, provides some guide. The simple linear trend of real labour income shown in Figure 3 shows a profile of permanent income that would be consistent with the way in which actual labour income evolved.^[21] According to life-cycle/permanent income theory, if households knew with certainty that income would evolve in this way, they would save more when current disposable household income was high relative to the long-run average growth path of labour income and save less when current disposable household income was low relative to the long-run average growth path of labour income. While it is unlikely that households would be able to predict the future course of labour income with anything like this degree of certainty, the broad trends in saving in Australia appear to correspond reasonably well with deviations of current income from this long-run trend in labour income.

A related point is made by EPAC (Whitelaw and Howe 1992) who argue that the slowdown in income growth since the mid 1970s (due to slower productivity and/or employment growth over much of the period) has been an important factor in explaining the general decline in saving in Australia. They point to the correlation between the domestic saving rate and the deviation in the level of per capita income from its long-run trend as evidence that consumers are slow to adjust consumption when income growth slows.

It also appears that there is considerable inertia with respect to consumers' response to short-term cyclical changes in income, a point also noted by EPAC (Whitelaw and Howe 1992). Permanent-income theory suggests that if short-term movements in income are dominated by transitory movements, current consumption will be invariant to short-run changes in income and saving will move closely with income. If there is uncertainty as to the permanence of any shock to income, even shifts in permanent income may not have an immediate effect on consumption spending and may increase saving in the short term Caballero (1990). Precautionary saving and the presence of liquidity constraints may also increase the short-term sensitivity of saving to changes in income. Edey and Britten-Jones (1990) find that short-run changes in the saving rate are well explained by changes in income.

4.3 Demographic Factors

In the life-cycle model, demographic factors such as age distribution, life expectancy, retirement age and labour market participation may all influence aggregate saving. Precautionary and bequest motives may reduce their influence somewhat, but demographic factors are still likely to be important.

There have been quite large shifts in the relative proportions of different life-cycle groups in the Australian population as well as marked changes in income profiles within some of these groups. The proportion of the population in the 45–64 year age group (predicted to be the main savers in the life-cycle theory) fell in the 1960s and 1970s, but has risen quite sharply over the past decade. The proportion in the older age groups, 65+, which the life cycle theory predicts to be dissavers, has been growing steadily.

Participation rates have also changed markedly. For example, Foster (1992) notes that participation rates of older workers have declined, further reducing the proportion of the population in those groups expected to be accumulating wealth for retirement, and increasing the retiree proportion of the population who are expected to be dissavers.

The large shifts in the demographic characteristics of the Australian population should be associated with quite pronounced movements in saving rates and it is likely that at least some of the fall in the saving can be attributed to these factors.^[22] There are however a number of factors that may mitigate against the importance of demographic factors. Precautionary and bequest motives, and illiquid wealth holdings, may each reduce the propensity of older households to run down assets in retirement to the degree predicted by the simple life-cycle model. Improved provision of social security may also have reduced the degree to which households accumulate wealth for retirement and the degree of subsequent dissaving.

Edey and Britten-Jones (1990) and Whitelaw et al. (1988) point to data from the Household Expenditure Survey which show that retirees reduce consumption in retirement in line with lower incomes, rather than maintaining consumption and dissaving at the rate suggested by the simple life-cycle model. If this is the case, there will be less variation in saving rates across different age groups and a weaker influence of demographic factors than the simple theory predicts.

4.4 Interest Rates

The life-cycle model ascribes an important role for real interest rates in the households intertemporal decision making process but the net effect on household saving is not clear. Edey and Britten-Jones (1990) argue that for Australia, the substitution effect should dominate because, in aggregate, the household sector receives a very small proportion of its income in the form of net interest payments. Empirical evidence is mixed, but overseas studies have generally found interest elasticities to be small and insignificant. Edey and Britten-Jones (1990) find similar results for Australia.

4.5 Indicators of Uncertainty

Uncertainty is difficult to measure at an aggregate level. Theory suggests that income volatility is an appropriate measure of uncertainty, but inflation and changes in the rate of unemployment are two measures which have also been used as indicators of uncertainty in other studies.^[23] Other proxies such as changes in hours worked, growth in employment, strike activity and labour market turnover could also be used.

Inflation may affect saving behaviour by causing typically risk-adverse households to save more for precautionary purposes. High inflation may be associated with financial or real shocks to the economy and with increased variability of income flows.

Fluctuations in the real economy are also likely to be associated with higher precautionary saving by risk-averse households. Increased unemployment, for example, may increase the expected variability of income flows of those households which are newly unemployed as well as those households who are at risk of becoming unemployed. Increased uncertainty may encourage households who may be affected, and particularly those who may face liquidity constraints, to increase saving in the short term.

We examine the effects of each of these possible influences on saving more formally in the next section.

Footnotes

Our measure of non-human wealth shown in Figure 2 includes financial assets such as household holdings of currency, deposits with financial institutions, government bonds, equities and superannuation assets as well as real assets such as dwellings and consumer durables. This measure is net of households' financial liabilities. [19]

See Simes and Horn (1986) and Lattimore (1994) for Australian evidence of non-human wealth effects. [20]

A linear, rather than log-linear, specification of the time trend was preferred because of the declining growth rates in labour income over the sample period. [21]

See Smith (1990), Lattimore (1994) and Bateman et al. (1991) for Australian evidence on the effects of demographic factors. [22]

See Andersen and Kennedy (1994) and Lattimore (1994). [23]