RDP 2007-08: The Effect of the Australian Superannuation Guarantee on Household Saving Behaviour 3. Data

This paper uses the confidentialised unit record dataset from the HILDA Survey (Release 3.0), a household-based panel study that began in 2001 from a reference population of all members of private dwellings in Australia. Wave 1 of the panel consisted of 7,682 households and 19,914 individuals. In Wave 2, conducted between 21 August 2002 and 19 March 2003, a special module collected comprehensive and detailed wealth data from each household. The key question that measures whether households receive compulsory pension contributions is: ‘Does your employer/business make contributions into a superannuation scheme on your behalf?’

Table 1 displays characteristics for five different samples to allow separate analysis of the effect of the reform for low-income and high-income households and to determine whether the presence of financial constraints is important, as theory would suggest. The first sample is the full sample, which includes all households where the household head is aged below 65 years and where someone in the household is employed and either receives or does not receive compulsory pension contributions.[10] The second sample is a subset of this full sample, including only households with below-median incomes, while the third sample includes only those with above-median incomes. The fourth sample is based on the full sample, but is limited to households where the household head indicated that they are financially constrained, following the definition in La Cava and Simon (2003).[11] The fifth sample includes all non-financially constrained households. I analyse financially constrained households separately since they are likely to have greater difficulty offsetting compulsory saving in pension accounts. Selecting households on this basis may also help to reduce unobserved heterogeneity, since households facing financial constraints may share saving behaviours that cannot otherwise be observed.

Table 1: Characteristics of Samples
Full sample Below-median income Above-median income Financially constrained Non-financially constrained
No compulsory pension contributions (per cent) 6.5 9.6 3.4 8.0 4.8
Median household disposable income ($) 49,705 33,398 70,145 46,520 54,239
Age of household head (years) 40 38 42 39 41
Makes voluntary super contributions (per cent) 36.8 24.3 49.4 32.1 42.2
Median net financial wealth ($) 73,630 37,060 127,250 49,830 108,000
Median net wealth ($) 223,100 106,734 390,500 155,613 333,700
Number of households 4,379 2,190 2,189 2,336 2,043

Source: HILDA Survey, Release 3.0

Around 6.5 per cent of households in the full sample do not receive compulsory pension contributions. This is broadly consistent with ABS estimates of the proportion of employees not covered by the system. By income, the proportion not covered ranges from 9.6 per cent for below-median-income households to 3.4 per cent for above-median-income households. The latter group is quite small, so results for the above-median-income sample should be interpreted with a greater degree of caution.

Two measures of net wealth are considered in this paper. The narrow definition of wealth includes financial wealth (‘net financial wealth’); while the broad definition also includes business equity and housing equity (‘net wealth’).[12] Since private businesses and houses are rarely traded, the accuracy of business and housing equity data is likely to be lower than it is for financial assets data, where prices are updated by financial markets. The narrow measure is equivalent to the net financial wealth measure used in the 401(k) literature by Poterba et al (1996). The two wealth measures are also constructed including and excluding pension assets so that the effect of receiving compulsory pension contributions can be broken into the effect on pension assets and the effect on other assets.

Unfortunately, it is not possible to measure directly the size of each household's compulsory pension account, C, since most households are also allowed to make voluntary contributions into the same account. Instead, I construct an estimate by taking the wages and salaries of employees who reported receiving pension contributions and extrapolate it back to 1992, when the Superannuation Guarantee was introduced, using data on average earnings by industry from the ABS. I then obtain the compulsory pension contributions by multiplying wages and salaries in each year by the contribution rate in that particular year, adjusting for the taxation of contributions. If the employees report that they were employed for less than 10 years (the length of time since the Superannuation Guarantee was introduced), I assume contributions have only been made for the number of years that they have been employed. I then cumulate the contributions into a stock using aggregate pension fund investment returns data adjusted for the taxation of investment income.

Job mobility is a potential source of measurement error in our estimates since we can only observe whether contributions were being made when respondents were surveyed during 2002/03. Nevertheless, as long as employees move between jobs where their contributions status remains unchanged, our measure remains sound. In the full sample, the household head spent, on average, 7.3 years in their current job (the median time was 4 years). However, the average period that the household head spent in the same occupation was 10.2 years while the median was 7 years. Assuming that jobs in the same occupation would be likely to offer similar working conditions, then job mobility may not be such a large source of measurement error. Furthermore, since the contribution rate gradually increased over the 1990s from 3 per cent to 9 per cent, wages in earlier years receive less weight in estimates of pension account balances.

3.1 Compulsory Pension Coverage

The bulk of the households not receiving compulsory pension contributions appear to fall into at least one of the exemptions. The first five rows of Table 2 show the percentage of the sample that fall within the exemptions. Another indicator of whether households are exempt is if tax is withheld from income. Employees who earn below the tax-free threshold do not need to have income tax withheld and are exempt, along with independent contractors and the self-employed, who are not remunerated through wages and salaries. Cumulatively, the exemptions appear to cover 77.5 per cent of the households where no working household members receive compulsory pension contributions.

Table 2: Explaining Compulsory Pension Coverage
Characteristics of the household head
No compulsory pension
Compulsory pension
Probit model of coverage
Per cent Cumulative
(per cent)
Per cent Cumulative
(per cent)
Marginal effect(c)
(per cent)
Labour income less than $450 per month 47.5 47.5   7.9 7.9   −3.6***
Self-employed 20.8 58.5   6.6 13.2   −5.1***
Under 18 and working fewer than 30 hours/week 1.1 58.8   0.0 13.2   −25.5*
Work in industry with exemption 5.6 62.0   5.3 17.9   −0.2
Income tax not withheld or not receiving wage/salary 59.9 77.5   8.2 19.2   −16.7***
No leave entitlements or impermanent contract 55.6 94.7   22.9 38.3   −10.2***
Fewer than 20 workers in workplace 71.1 96.5   38.1 55.7   −0.9*
Service industries with low coverage rates 52.1 97.8   26.8 63.6   −1.7***
Pseudo R2             32.6

Notes: *** and * represent significance at the 1 and 10 per cent levels, respectively, for the test of the underlying coefficient being 0.
(a) 6.5 per cent of full sample
(b) 93.5 per cent of full sample
(c) The marginal effect is for discrete change of dummy variable from 0 to 1

Source: HILDA Survey, Release 3.0

The remaining households that do not receive compulsory pension contributions may have jobs in the underground economy. These households represent 1.4 per cent of the full sample, consistent with estimates of the size of the underground economy. Drawing on research by the ABS (2003a), I can identify several factors that may be correlated with such jobs, including: no leave entitlements or an impermanent contract; small workplaces; or service industries where remuneration outside the tax system may be more prevalent.[13] Cumulatively, 97.8 per cent of the households not receiving compulsory pension contributions can be explained by the exemptions or the factors correlated with jobs in the underground economy.

A probit model confirms that most of the factors in Table 2 significantly reduce the probability that the household will receive compulsory pension contributions. The only factor that is statistically insignificant is ‘work in an industry with exemption’. This result may be due to employment-by-industry data not being sufficiently disaggregated to distinguish the exemptions accurately. Even so, many households not receiving compulsory pension contributions have similar characteristics to those who do: 64 per cent of the households receiving contributions share at least one of the characteristics listed in Table 2. This can give us some confidence that households not receiving compulsory pension contributions represent a reasonable control group, since they do not appear to be intrinsically different from many of the households receiving contributions.


This excludes households for which no definitive answer was given to the question of whether employer contributions were made. The household head is defined as the most important provider of income. A hierarchy of characteristics is used to determine the household head: i. income; ii. labour income; iii. labour force status; and iv. age, with characteristics lower down the hierarchy only used in case of a tie between household members. [10]

The financially constrained are defined as responding ‘yes’ to any of the following questions: having difficulty raising $2,000 in an emergency (for example, would need to borrow funds, sell an asset or could not raise the funds); difficulty paying utility bills on time; difficulty paying mortgage/rent on time; pawned or sold something; went without meals; was unable to heat home; asked for financial help from friends or family. [11]

Housing/business equity is defined as housing/business assets minus housing/business debts. [12]

As measured by the ABS, the industries with lower employer superannuation coverage are: agriculture, forestry & fishing, construction, retail trade, hospitality, cultural & recreational services and personal & other services. [13]