RDP 2007-08: The Effect of the Australian Superannuation Guarantee on Household Saving Behaviour 4. Empirical Results

4.1 Household Wealth

Table 3 presents the results for the estimated median effect of receiving compulsory pension contributions on the wealth-to-income ratio (Equation (1)). For the full sample, receiving contributions significantly increases the net financial wealth-to-income ratio by 20.3 per cent of income. When I disaggregate this into the contribution from pension assets and other assets, the effect is mainly due to an increase in pension assets of 15.8 per cent of income. Households do not appear to be offsetting the increase in pension assets by reducing other financial wealth, since the point estimate for the effect on non-pension assets is positive (4.4 per cent) and statistically insignificant. When business and housing equity are added to obtain a measure of total net wealth, the results are not statistically significant but the effect is still economically sizeable, at 46.7 per cent of income.

The results for the sub-samples show that the effect of receiving contributions on the financial wealth-to-income ratio is positive and generally significant at the 10 per cent level. For below-median-income households, the effect on financial wealth is 21.1 per cent of income and appears to be dominated by an implied effect on pension assets of 17.4 percentage points. When the net wealth-to-income ratio is used, the results are insignificant for the sub-samples. While it is likely that net wealth is measured less accurately than financial wealth, reducing the precision of our estimates, the effect of receiving contributions on net wealth is economically significant for all the samples. The point estimates are larger for below-median-income and financially constrained households than when net financial wealth is used, however, the differences are not likely to be statistically significant.

While the results in Table 3 suggest that compulsory pension contributions have a positive effect on net wealth, they only provide an imprecise estimate of the magnitude of the effect. More precise estimates of the marginal effect of a dollar of compulsory pension assets on household wealth can be obtained by estimating Equation (2), with the results presented in Table 4.^[14] Overall, the results suggest that marginal increases in compulsory pension account balances largely flow through to increases in household wealth. For the full sample, an extra dollar in compulsory pension accounts increases net financial wealth by 91 cents, suggesting an offset of only 9 cents through reductions in other assets. This appears to occur mainly through reductions in non-pension financial wealth, which falls by a statistically significant 19 cents; it should be noted that the equivalent coefficient in Table 3 was positive, though insignificant. When the broadest measure of wealth is used, there does not appear to be evidence of any offset, with household saving increasing by 133 cents for every extra dollar in compulsory superannuation. While this point estimate implies that pension assets increase by more than a dollar in response to an extra dollar of compulsory pension assets (partly offset by a reduction in non-pension financial wealth), this estimate is not significantly different from 1. In other words, the results for the broader measure of net wealth are once again imprecisely estimated and should be treated with a high degree of caution.

The results for the sub-samples show that the effect of an extra dollar in compulsory pension accounts on net financial wealth is largest for financially constrained households, whose financial wealth increases by 93 cents for every extra dollar in their compulsory pension accounts. In contrast, the effect on non-financially constrained households is smaller at 85 cents and is not statistically significant (the difference between the two estimates is also likely to be statistically insignificant). This is consistent with our expectations, since financially constrained households would have less opportunity to reduce holdings of other assets to offset compulsory saving. The marginal effect of compulsory contributions on net financial wealth for above-median-income households is 83 cents, close to the effect for below-median-income households of 77 cents. In particular, above-median-income households (and non-financially constrained households as well) appear to have a larger stock of non-pension saving with which to offset compulsory contributions. Their financial wealth excluding pension assets falls by 30 cents for every extra dollar in compulsory pension accounts, while the non-pension assets of below-median-income households are not significantly affected. Interestingly, compulsory contributions appear to have more than a dollar-for-dollar effect on the pension assets of above-median-income households (and those households who are non-financially constrained). This result may be due to high-income earners making voluntary pension contributions to take advantage of their concessional tax treatment.

Overall, it would be reasonable to characterise the results in Table 4 as suggesting that household wealth increases by around 70 to 90 cents for every extra dollar in compulsory pension accounts, with the effect most pronounced for financially constrained households. This is broadly consistent with our expectations as outlined in the introduction.

4.2 Voluntary Saving for Retirement

Receiving compulsory pension contributions appears to have a positive influence on the propensity of households to voluntarily save for their retirement in pension accounts. The results of a probit model on whether households make voluntary contributions (Equation (3)) are presented in Table 5. For all the samples, the probability of making a voluntary contribution increases when households receive compulsory pension contributions and the effect is statistically significant at the 5 per cent level. In the full sample, the probability of making voluntary contributions is 19 per cent higher if the household receives compulsory contributions.

It is somewhat surprising that the effect on voluntary retirement saving is positive, since we would expect households to first offset compulsory contributions in pension accounts by reducing voluntary contributions. It is possible that compulsory pension contributions lead to higher voluntary pension contributions by highlighting the importance of retirement saving and making it more convenient. This result is similar to findings for the United States by Madrian and Shea (2001), where households are more likely to voluntarily save in pension accounts if they are automatically enrolled by their employer.^[15]Nevertheless, households could still offset their retirement saving by reducing non-pension wealth, such as housing equity or other investments.

To quantify the effect on voluntary retirement saving, Table 5 also presents the results of a tobit model where the regressand is the size of voluntary contributions as a per cent of income. The marginal effect of compulsory contributions is to increase voluntary contributions by around 1.5 per cent of income in the full sample. The marginal effect is estimated to be around 1 per cent of income for below-median-income and financially constrained households, compared with closer to 2 per cent for above-median-income and non-financially constrained households, implying that financial constraints do, to some extent, restrict the ability of households to save more for retirement.

The magnitude of the effect across the samples appears small but is economically significant, particularly if the extra 1.5 per cent of saving each year occurs throughout the working life of the household. Under conservative assumptions regarding investment returns on assets in these accounts, over a 40-year working life, this extra saving would cumulate after tax so as to increase the wealth of the household by 75 per cent of pre-retirement income.^[16] Assuming a replacement rate of 40 per cent of pre-retirement income, this extra saving would fully fund around 2 years of retirement. Over a 25-year retirement period, this would boost the replacement rate by around 5 percentage points.

4.3 Retirement Intentions

Table 6 presents results for the estimated effect of compulsory contributions on the average retirement intentions of households where at least one household member is still working and aged between 45 and 64. First, I find that households who receive compulsory contributions are no more likely to report an intended age of retirement (Equation (4)). Of the households where at least one member specified an intended age of retirement, receiving compulsory contributions did not have a significant effect on the average intended age of retirement (Equation (5)). These results are broadly consistent whether or not households were below the median income or were financially constrained.

These results suggest that compulsory pension contributions do not have a significant effect on the retirement intentions of households. This is somewhat surprising given the importance of these contributions to retirement incomes. It is possible that the Superannuation Guarantee is affecting retirement intentions in several ways that could possibly be offsetting each other. For instance, the increase in net wealth due to compulsory pension accounts could bring forward the retirement age of households, assuming leisure is a normal good. However, these households may decide to delay retirement if the Superannuation Guarantee makes them aware that they need to save more to achieve their desired standard of living in retirement.

There are some caveats that should be taken into account when interpreting this result. In particular, it should be noted that household members aged over 45 in 2002/03 would have only been exposed to compulsory pension contributions for a fraction of their working life, while younger household members will be exposed to the system for their entire working life. Therefore this result may not be so relevant to the wider working population. Also, it is possible that retirement intentions are a poor indicator of when households will actually retire, as many household members may not have planned their retirement in detail at this stage.

4.4 Robustness Tests using Matching Estimators

An alternative way of estimating the effect of compulsory pension contributions on household saving behaviour is to use matching estimators. The technique is analogous to using the covariates, X_i, in the models in Section 2 to find a ‘financial twin’ for each household not receiving compulsory contributions from the group of households receiving contributions.^[17] After obtaining the closest match for each household not receiving compulsory contributions, the effect of the Superannuation Guarantee on household wealth, retirement saving and retirement intentions can be calculated.

The results are presented in Appendix A, Tables A2 and A3, and suggest that the regression results presented above are broadly robust to the use of this alternative estimation technique. In particular, compulsory pension contributions appear to have a sizeable effect on household net financial wealth and net total wealth. Households receiving compulsory contributions also appear more likely to voluntarily save for retirement in pension accounts. The only significant difference with the regression results above is that households receiving contributions intend to retire later when the matching estimator is used.

Footnotes

See Appendix A, Table A4 for the complete set of estimates from the net financial wealth model for the full sample. [14]

Covick (2002) suggested that means-testing arrangements on social security programs could also help to explain why some households may choose to voluntarily save more for retirement in response to a compulsory saving scheme. [15]

Assuming that nominal labour income grows by 4 per cent, nominal returns to pension funds average 7 per cent and that contributions and returns are taxed at 15 per cent. [16]

See Abadie et al (2004). [17]