RDP 2014-03: Household Saving in Australia 4. Time Series Analysis

In the previous section we considered how household characteristics relate to median saving behaviour in the cross-section. We now focus on the drivers of the rise in the saving ratio between 2003/04 and 2009/10. To do this, we look at changes in households' propensity to save using the median regression model from Section 3; we also decompose the total change in the mean saving ratio (the concept of saving reported in the national accounts) into changes in households' propensity to save and changes in household characteristics.

4.1 Changes in the Median Saving Ratio

The last column of Table 1 in Section 3 shows the change in model coefficients between the 2003/04 Survey and the 2009/10 Survey. We interpret changes in these coefficients, where they are statistically significant, as indicating changing preferences regarding saving for those households with the corresponding characteristics. As noted in Section 1, however, since we cannot directly measure household preferences, other interpretations of the data are possible.

Income

There is no change in the coefficient on deviations of current relative to permanent income between the two surveys. There is a significant change in the coefficients on education, however. Relative to high school educated households, more highly educated households significantly increased their propensity to save between 2003/04 and 2009/10. If we interpret education as a proxy for future income expectations, this suggests that higher-educated households downgraded their income expectations between 2003/04 and 2009/10; this implies current income being high relative to permanent income, which would lead households to spend less and save more.

Financial constraints

The propensity to save for financially constrained households did not increase between 2003/04 and 2009/10. As such, our results do not support the hypothesis that tighter credit constraints played a significant part in the rise in household saving.

Variables related to precautionary motives

Single-parent households and households who received more than 20 per cent of their income from government payments increased their propensity to save between 2003/04 and 2009/10 compared with reference households, suggesting that these households were less resilient than other households to changes in financial circumstances.

Self-funded retirees and those earning at least 20 per cent of their income from investments – that is, those households most exposed to movements in asset prices – also increased their propensity to save between 2003/04 and 2009/10, suggesting a reaction to the large fall in asset prices that occurred during the financial crisis.^[10]

Variables related to wealth

The wealth effect for households aged 50 to 64 years fell significantly between 2003/04 and 2009/10, while the negative effect on saving of owning a home outright also fell for households aged 65 and over. Lowe (2011) suggests that weakening wealth effects for older households could be due to the slower growth in the value of dwelling assets in the period leading up to 2009/10 and/or wealthy households losing liquid assets in the financial crisis and saving more to rebuild their wealth.

The negative effect on saving of a high gearing ratio also fell between 2003/04 and 2009/10, with the fall for young households being statistically significant. This suggests that households may have adopted a more prudent attitude to debt between 2003/04 and 2009/10, and accords with other data sources that suggest households have increased their voluntary mortgage repayments over the past few years, aided by lower interest rates (RBA 2012).

4.2 Changes in the Mean Saving Ratio

Using the same model, but applied to the mean, the model-implied mean saving ratio in year i can be expressed as

where is a vector of the averages of variables used in the saving model in year i, including the constant term, and β_i is a vector of the coefficient terms associated with the variables in in year i. Given this, we can express the change in the mean saving ratio as

where year 1 represents 2003/04 and year 2 represents 2009/10. That is, the change in the model-implied mean saving ratio can be decomposed into changes in model parameters and changes in population characteristics. This follows the method introduced by Blinder (1973) and Oaxaca (1973).^[11],^[12]

This decomposition enables us to estimate the roles that population characteristics and model parameters have played in the rise of the household saving ratio separately. The results suggest that changes in population characteristics played virtually no role in the increase in the saving ratio between 2003/04 and 2009/10, with changes in model parameters dominating (Figure 7).

Figure 7 shows the total contribution of changes in model parameters to the change in the model-implied mean saving ratio (the red bar on the left), as well as the contribution from variables related to: education, precautionary motives (split into those related to incomes and those related to assets), and wealth.^[13] With the exception of wealth, all subgroupings are statistically significant at the 5 per cent level; the wealth subgrouping is statistically significant at the 10 per cent level. Not shown are the constant (negative although not statistically significant), other variables from Section 3 that were not statistically significant, and the combined effect of other variables and controls not otherwise discussed (again not statistically significant). The combined effect of all these together is, however, statistically significantly negative.

The results presented in Figure 7 are consistent with the results from the median analysis in Section 4.1.

• Higher-educated households increased their propensity to save between 2003/04 and 2009/10, with this increase largest for the most highly educated households. Given our interpretation of education as a proxy for permanent income, or equivalently for expectations regarding future increases in income, the rise in saving for higher-educated households suggests a downward reassessment by these households of their future income prospects.
• The propensity to save rose for those households with attributes suggestive of less secure income or vulnerability to asset price shocks, which suggests a greater degree of risk aversion, or a greater degree of risk, for households with these characteristics.
• Finally, wealthy households and those with high debt levels (included in the wealth grouping) tended to increase their propensity to save between 2003/04 and 2009/10, suggesting an effort to rebuild wealth after the effects of the financial crisis and changed attitudes to debt.

Saving ratios associated with a number of other variables discussed in the median analysis were not found to statistically significantly change between 2003/04 and 2009/10, including the risk of being unemployed and the credit constraint variables. Regarding other variables:

• By construction, the deviation of current income from permanent income plays no role in the modelled change in the saving ratio, since the average deviation of temporary income from permanent income is zero in both survey years.
• With the exception of pre-retirement-aged households (included in the vulnerable to asset price shocks subgrouping), the change in the age effects are not statistically significant.

Overall, the results from the median and mean time series analysis are consistent with a number of factors driving the increase in household saving between 2003/04 and 2009/10. The rise in saving for those groups judged to be vulnerable to income or asset price shocks suggests that precautionary motives played a role, with households observing events overseas, as well as rising unemployment and declines in asset prices domestically, and judging the world to be a more risky place than previously thought. Related to this, the rise in saving for those with high debt levels suggests that households adopted a more prudent attitude towards debt over this period, while the rise in saving for higher-educated households suggests a downward reassessment of expected future income prospects for these households. Finally, the rise in saving for wealthy households suggests a reassessment of expected future capital gains and a desire to rebuild wealth, with declines in asset prices following the global financial crisis both reducing wealth immediately and reminding households that asset prices can fall as well as rise.

However, as discussed earlier, since we cannot directly measure household preferences, we can only draw inferences based on which household groups changed their propensity to save, and other interpretations of the data are possible.

Footnotes

Note that self-funded retirees are likely to dissave more than suggested by our results. As discussed in Section 2.1, in our dataset we cannot separately identify capital draw-downs from investment earnings for self-funded retirees. As such, some of the income attributed to self-funded retirees is actually dissaving from their accumulated assets. [10]

As noted, the model used here is very similar to the model for the median saving ratio in Section 3, except that it is estimated by least squares. As such, the model is of conditional mean saving ratios rather than conditional median saving ratios. We also drop the interaction terms between age and the wealth-related variables. See Table B4 for output from the least squares regression. [11]

Appendix D presents a quantile decomposition of the change in the saving ratio for each percentile of the saving ratio distribution. [12]

The education category includes dummy variables for households with TAFE/certificate and university education; the less secure income category includes dummy variables for unemployed, pensioner, single-parent, lower-skilled, non-English-speaking migrant households and households that rely on the government for income; the vulnerable to asset price shocks category includes dummy variables for self-funded retirees and pre-retirement-aged households; and the wealth category includes the wealth-to-income ratio, home ownership dummies and the gearing ratio. [13]