Number 33: Structure and distribution of household wealth in Australia: cohort differences and retirement issues

Executive summary

In 2002, the first large-scale survey of household wealth in Australia since World War I was undertaken as part of Wave 2 of the Household, Income and Labour Dynamics in Australia (HILDA) survey. The wealth data obtained in the survey (from 7,245 households) covered all the main components of asset portfolios and debts. The data were found to match satisfactorily with national aggregate statistics available from the Australian Bureau of Statistics and the Reserve Bank of Australia.

The aim of this paper is to analyse the 2002 HILDA wealth module data under a number of headings including:

  • the structure and composition of household wealth and debt and whether this has changed in recent years
  • the distribution of wealth, including among and within age cohorts and the most and least wealthy cohorts
  • retirement issues, focusing in particular on the capacity of pre and post-retirement cohorts for self-funding during retirement
  • the wealth and debt levels of vulnerable groups in society, especially income support recipients and lone parents
  • the main factors (demographic, educational, income related, and so on) which determine levels of wealth and debt.

Analysis of the HILDA data shows that asset holdings are heavily concentrated in the hands of older households—those within 20 years of retirement and those 10 to 15 years post retirement. This distribution is due mainly to the fact that asset levels depend on the length of time spent saving and benefiting from the effects of compound interest. It is also a consequence of government policy and initiatives such as the Superannuation Guarantee and generous tax concessions to encourage retirement savings.

Superannuation holdings are increasing rapidly and are now more widely distributed than in the past. However, the evidence in this report shows that most households within 20 years of retirement are likely to be partly reliant on the Age Pension for their retirement income. Current government policy is tackling this problem by changing incentives affecting both the age at which people choose to retire and their likelihood of doing some paid work during retirement. It remains to be seen how effective these changes will be in counteracting the evident desire of most Australians to retire before age 65. A serious underlying problem is that most working-age people continue to underestimate the savings they will need to maintain their current lifestyle in retirement.

The main assets types of Australian households are housing, equities and superannuation. Given the volatility in the value of these assets in recent times, it is probably mistaken to believe that household wealth is fairly stable. In the context of an ageing population, we need a better understanding of the dynamics of wealth, particularly for those in the retirement and pre-retirement cohorts. This means it will be important to measure household wealth and assess the causes and consequences of change more frequently than in the past.

[ Return to Top   Return to Section ]

1. Introduction

1.1 Aims

This report is intended to contribute to an understanding of the structure and distribution of wealth in Australia. More particularly, it relates to the extent to which mature-age Australians now have, and may have in the future, a capacity for financial self-reliance during retirement.

The aim of this report is to analyse the wealth module in the 2002 Household Income and Labour Dynamics in Australia (HILDA) survey in order to address the following issues.

Structure/composition of wealth and debt

Issue 1

What are the main components of wealth and debt? What are the relative shares of financial assets and non-financial (particularly housing) assets?

Issue 2

Has the level and composition of household wealth changed in recent years? If so, to what extent have increases been due to possibly transient increases in housing values?

The distribution of wealth

Issue 3a

What is the overall distribution of wealth?

Issue 3b

How are wealth and indebtedness distributed among age cohorts? How wide is the gap between the most and least wealthy cohorts? How is wealth distributed within cohorts?

Retirement issues

Issue 4a

Retirement—focusing on the older age cohorts (those aged 45 years and over), what is their capacity for self-funding in retirement?

Issue 4b

Is there a relationship between intended age of retirement and levels of wealth and debt?

Vulnerable groups in society

Issue 5

How do vulnerable and ‘at risk’ groups fare in regard to wealth? In particular, what are the assets and debts of income support recipients? What is the wealth situation of lone parents?

Determinants of wealth

Issue 6

What are the main factors (demographic, educational, income related, and so on) which determine levels of wealth and debt?

1.2 Why wealth matters and why it makes sense to think of it as a long-term ‘stock’

Before addressing these issues directly, it may be useful to ask why wealth matters. How does it contribute to a household’s economy and quality of life and what are the implications of seeing wealth as a long-term ‘stock’?

Wealth confers economic security, and this is a high priority for many people. It enables a household to tide over bad times due to, for example, unemployment or ill health, when the normal flow of earned income is reduced or cut off. Wealth also enables a household to gain access to credit so it can borrow either to tide over bad times, or invest for the future by, for example, paying for education, or buying property, shares or a business. Wealth also directly generates income both in cash and in kind; for example, shares and superannuation holdings directly generate cash income. Equally valuably, owner-occupied housing or art works or other collectibles in the home provide benefits in kind. They contribute to a household’s quality of life and standard of living broadly defined. In the context of this report, a key aspect of wealth is that it can provide security and even comfort in one’s retirement years.

It makes sense for individuals and households to think about wealth in the long term, not just the short term. In economic or accounting terms wealth is a ‘stock’, whereas income is a ‘flow’. Households acting rationally think about building up their wealth over members’ working lives. Then, in retirement, members use their accumulated savings to enjoy a satisfying standard of living and quality of life.

A corollary of the long-term nature of wealth planning is that it is sometimes sensible for households to incur debts so they can invest and build up their assets in the long term. In the short term, a household might rationally choose to reduce its net worth (assets minus debts) in order to achieve long-term gains. A household approaching retirement may, for example, borrow against the equity in its home to buy shares or managed funds, knowing that it will consequently, for some years to come, have both high asset levels and high debts. We shall see throughout this report that high assets and high debts go together. The more assets you have, the better your credit rating is and the more you are likely to borrow. Debts and investments are, of course, often the same thing.

Partly because of these considerations, we use a range of wealth measures in this report, rather than a single summary measure. We consistently use net worth as a summary measure, but we also report household assets and debts separately in most tables, and often components of financial assets and non-financial assets. Financial assets include superannuation, investments in shares and other equities, cash-type investments and bank accounts. Non-financial assets include housing, businesses and vehicles.

1.3 Australian data and research on wealth prior to HILDA

In 1915, faced with increasing wartime expenditure, the Australian Government used the Census to measure the wealth of the five million or so Australians living in the country at the time. As it turned out, nearly 90 per cent of all wealth was owned by the richest 10 per cent of the population.

The wealth module included in the second wave of HILDA in 2002 is the first large-scale survey of household wealth (n=7,245 households) conducted since the 1915 Census. The module was designed jointly by the Reserve Bank of Australia (RBA) and the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The questions covered all main financial assets, including bank accounts, superannuation and shares, and all main non-financial assets, including housing and businesses, together with the main categories of debt.

Because this was a household survey, rather than an estimate of national aggregate wealth of the kind published regularly by the Australian Bureau of Statistics (ABS), it enables us to focus on distributional issues and cohort differences—that is, the differing asset portfolios of richer and poorer households and of different age cohorts.

The main data series available from the ABS do not deal with distributional issues. They provide quarterly estimates of national aggregate household wealth (ABS cat. no. 5232.0).1 The value of some non-financial assets, principally housing but not unincorporated businesses, are estimated in the National Accounts (ABS cat. no. 5204.0). In general terms, government agencies (such as the ABS and the RBA) do not measure household sector assets directly, but rather treat them as residuals after subtracting business sector assets (about which much is known) from national estimates of wealth. Even so, while existing sources probably provide accurate estimates of changes in household wealth over time, they tell us little about distributive issues.

Most previous estimates of the distribution of household wealth have been based on inferring stocks (assets) from flows (income). For example, income from business ownership and income from share dividends have to be declared on tax returns and can be used to estimate asset values. The National Centre for Social and Economic Modelling (NATSEM) has used this approach to provide household estimates (Baekgaard 1998; Kelly 2001, 2003), as has the ABS (Robertson, Grandy & McEwin 2000; Northwood, Rawnsley & Chen 2002).

Apart from the 1915 Census, there appears to have been just one small-scale Australian survey of household wealth conducted in 1967 (Podder & Kakwani 1976). This was valuable but it omitted farmers and, partly due to this omission, may have understated inequality of net worth (Kelly 2001).

The next chapter gives an overview of the HILDA wealth module and the main measures used. HILDA’s wealth estimates are compared with the ABS and RBA national aggregates.

[ Return to Top   Return to Section ]

2. Methods

2.1 The HILDA wealth module 2002

A wealth module was included in the second wave of HILDA in 2002. Described in more detail in Watson and Wooden (2002), HILDA is a household panel survey conducted under contract to the Australian Government Department of Families, Community Services and Indigenous Affairs (FaCSIA) by the Melbourne Institute at the University of Melbourne. It began in 2001 with a large, national, representative sample of households and involved personal interviews with all household members aged 15 and over. In Wave 1, interviews were obtained from 7,682 households, which represented 66 per cent of all households identified as in scope. This in turn generated a sample of 15,127 persons eligible for interview, 13,969 of whom were successfully interviewed.

The survey's coverage is extremely broad, but focuses on household structure and formation, income and economic wellbeing, and employment and labour force participation. Each year a special module of non-core questions is added. In Wave 1, it was appropriate to devote the module to personal and family history. In Wave 2, the topic was wealth, with the module being funded in large part by the RBA.

In 2002 all responding households from Wave 1 were re-contacted. Interviews were again sought with all household members aged 15 or over, including persons who did not respond in Wave 1, as well as any new household members. In total, interviews were completed with 13,041 persons from 7,245 households. Of this group, almost 12,000 were respondents from Wave 1, which represented almost 87 per cent of the Wave 1 individual sample. Like all surveys, the HILDA survey has sampling errors—differences between the sample's characteristics and population Census characteristics. The data include weights to adjust for these biases.

2.2 Measuring wealth

The HILDA wealth module was designed jointly by the RBA and the Melbourne Institute. Most of the questions about assets and debts were asked at the household level and answered by one person on behalf of the entire household. The questions covered housing, incorporated and unincorporated businesses, equity-type investments (for example, shares, managed funds) and cash-type investments (for example, bonds, debentures), and vehicles and collectibles (for example, art works). However, individuals were asked some questions about assets and debts, including superannuation, bank accounts, credit cards, student debt (Higher Education Contribution Scheme—HECS) and other personal debt. Respondents were asked to give exact dollar amounts when answering all questions. However, bands were offered to those who could not provide an exact estimate of their superannuation holdings—a particularly difficult topic. (Appendix A presents a diagram showing how the components of wealth measured in HILDA have been aggregated.)

Wealth is not easy to measure in surveys and, when attempted overseas, has been associated with high item non-response rates and substantial underestimates in measuring aggregate national wealth, if the National Accounts are taken as a benchmark (Juster, Smith & Stafford 1999). This last result is largely due to the fact that the wealthiest 2 per cent or so of the population, who own a vastly disproportionate share of the wealth, are invariably underrepresented in surveys.

The HILDA survey was not immune from these difficulties. It achieved response rates of more than 90 per cent for almost all components of wealth, but 39 per cent of households could not provide information about at least one component. Not surprisingly, many could not specify their superannuation or business holdings. So we were only able to directly compute net household wealth for 61 per cent of responding households. Statistical imputations of missing data were undertaken by the RBA. This was essential, since to omit cases with missing data would have introduced a bias against larger households in which it is harder to get every eligible respondent to participate. So the HILDA files used in this report now contain complete wealth data for all 7,245 households interviewed at Wave 2, with flags to indicate where imputations have been made.

2.3 Benchmarking HILDA wealth estimates: comparisons with estimates in ABS and RBA publications

In assessing currently available data it is usually helpful to have in mind what an ideal data set would look like. In measuring wealth, then, we ideally want to measure the market value of all assets and debts. This is what HILDA tried to do. Alternative measures exist and are, in some cases, used by government agencies as proxies for market value. An example is use of the book (tangible assets) value of business assets. This measure has the advantage of being reliable in the sense that it is replicable with a low margin of error. But, if it is substantially different from market value (which it probably is—see below), then it is not a valid measure; that is, it is not measuring the 'right' thing, but something else. Officials in government agencies who use proxy measures are presumably aware of the validity limitations of such measures, but the measures are quite often cited by others as if they were indisputably valid. A key point about measurement is that measures cannot be valid unless they are reasonably reliable, but they can be totally reliable and have no validity at all.

To benchmark the HILDA data we compare its estimates of household assets, debts and net worth with the national aggregates available from the main ABS and associated sources. The main published sources are:

(1) the ABS Australian System of National Accounts (cat. no. 5204.0), which needs to be read in conjunction with the ABS Financial Accounts (cat. no. 5232.0)

(2) RBA Statement on Monetary Policy (various dates).2

As a first step in benchmarking, Table 1 shows which components of financial assets, non-financial assets and debts are reported in HILDA and in the government sources.

Table 1: Aggregate wealth sources, summary of differences in scope
Asset type HILDA ABS RBA
Financial assets      

Deposits

Bonds etc.

Equities

Unfunded superannuation

X

Pre-paid insurance premiums

X X
Non-financial assets      

Vehicles

Other consumer durables

X

Housing and property

Business assets

X

Collectibles

X X
Debts      

Housing debt

Business debt

Student debt (HECS)

Credit card debt

Other personal debt

Note: √ indicates that the component is included by HILDA, ABS or RBA, while x indicates it is not.

Differences of inclusion and exclusion among the different sources mean that comparisons are not straightforward. One immediate difference is that the ABS and RBA include the assets and liabilities of non-profit organisations in the 'household balance sheet', whereas HILDA data relate strictly to households.

Now consider the treatment of financial assets. We see that the ABS and the RBA only differ in how they treat unfunded superannuation and pre-paid insurance premiums—the RBA excludes them whereas the ABS does not. Conceptually the HILDA survey falls between the two, including unfunded superannuation but excluding pre-paid insurance premiums. (It should be noted that the sums involved here are quite small.) Turning to non-financial assets, Table 1 indicates that HILDA measures collectibles whereas the government sources do not. On the other hand, HILDA does not measure the value of consumer durables (except vehicles), whereas the government sources do. Both HILDA and the ABS value the assets of unincorporated businesses owned by households, but the RBA leaves them out.

Let us now compare dollar amounts for assets and debts estimated by HILDA and the government sources. Estimates for HILDA and the ABS relate to an average of the September 2002 and December 2002 quarters, which is when most of the HILDA interviewing was conducted. The RBA data are for the December quarter. Note that the HILDA estimates of mean asset and debt values, which were obtained on a household basis, were multiplied by the number of households in the country to obtain comparisons with the national aggregates provided by the government sources.

Financial assets

First, we consider financial assets. As noted above, conceptually HILDA falls between the two government sources by including unfunded superannuation but excluding pre-paid insurance premiums. It transpires that, empirically, the HILDA estimate also lies between the government estimates:

HILDA: $1,125 billion

ABS: $1,237 billion

RBA: $1,084 billion

If we adjust the HILDA data by adding the ABS estimate of pre-paid insurance premiums—just over $28 billion—we find that the HILDA estimate is about 93 per cent of the ABS estimate. Thus, as expected, HILDA understates the volume of financial assets. The size of that understatement, however, is relatively modest.

Non-financial assets

Next, we consider non-financial assets, the area in which it is most difficult to make well-matched comparisons. Housing is by far the biggest component, followed by business assets. At first examination, the three sources seem far apart on housing:

HILDA: $1,932 billion

ABS: $1,597 billion

RBA: $2,252 billion

In large part the explanation for these discrepancies lies in different methods of measurement/estimation. The HILDA survey asked respondents who completed the Household Questionnaire (nearly always the household reference person or his/her partner) to estimate the market value of their property if sold today. There is strong Australian evidence that householders can estimate their own property values with reasonable accuracy—on average they get within 3 per cent of the 'correct' valuation as determined by professional real estate valuers (Yates 1991). The ABS adopts a quite different approach. A perpetual inventory model is used to estimate the dwelling stock, allowance is made for new building activity, and values are obtained from the Housing Industry Association (HIA)/Commonwealth Bank of Australia (CBA) house price series (Northwood, Rawnsley & Chen 2002). The RBA does things somewhat differently: it '… splices together the quarterly HIA/CBA median price series and the Real Estate Institute of Australia (REIA) median price series for each state' (Northwood, Rawnsley & Chen 2002, p. 53). The different methods used by the two government organisations produce widely divergent results and there is some evidence that, over time, the gap is getting wider. Northwood, Rawnsley and Chen (2002) show that, between June 1994 and June 2000, the gap increased from about 25 per cent to about 50 per cent with the RBA estimate always being higher. It is not possible for us to determine whether the ABS or the RBA estimates are the more accurate. What can be said is that the ABS estimates are consistently the lowest, by a large margin, of all the main estimates produced in Australia (Northwood, Rawnsley & Chen 2002, p. 53). They are lower than estimates from the Treasury, the REIA, the HIA/CBA, and the RBA.

So where does HILDA stand on housing and property values? The figures above indicate that its estimate is about 86 per cent of the RBA estimate (which, however, was for the December quarter rather than an average of the last two quarters of 2002) and 121 per cent of the ABS estimate.

Business assets

Next, we consider business assets:

HILDA: $339 billion

ABS: $151 billion3

RBA: same data as ABS

Again, the HILDA and ABS figures are far apart. The main reason appears to be conceptual. HILDA sought to measure the market value of unincorporated business by asking the household respondent (described as the household member who was best informed about the household's wealth) to estimate the value of the business if sold today. The RBA uses the book value (that is, the tangible asset value) of businesses, which is not a valid proxy for market value. However, it has been estimated that typically—but with huge variations—the book value of a business tends to be about half the market value (Webster 2000). But it should be noted that estimates of the ratio of book value to market value are much less precise for small and unincorporated businesses not listed on the stock exchange than they are for listed companies. In the case of listed companies, market value corresponds to share prices. No such handy valuations exist for unincorporated entities. A final small difference between HILDA and the ABS is that HILDA did not distinguish between the financial and non-financial assets of businesses and just classified all business assets as non-financial. The ABS figure above relates only to non-financial assets.

Cars and other household vehicles

Next we consider, under the heading of non-financial assets, valuations of cars and other household vehicles:

HILDA: $143 billion

ABS: $50 billion

RBA: same data as ABS

HILDA's estimate of vehicle values at $143 billion is nearly three times higher than the ABS estimate. It seems that, although HILDA respondents may have slightly misestimated the value of their own vehicles, the discrepancy may be due to the ABS applying high discount rates to vehicles and, in effect, attributing zero value to vehicles still on the road.

Housing, businesses and vehicles are the three types of non-financial assets on which HILDA data and the government sources can be compared (see Table 1). In any event the value of other non-financial assets is not significant to the aggregate. Summing the estimates for non-financial assets we get the following results:

Major non-financial assets = housing + household business + vehicles

HILDA: $1,932 billion + $339 billion + $143 billion = $2,414 billion

ABS: $1,597 billion + $151 billion + $50 billion = $1,798 billion

RBA: $2,252 billion + $151 billion + $50 billion = $2,453 billion

On this basis it appears that the HILDA estimate is 98 per cent of the RBA figure and 134 per cent of the ABS figure. The differences, as explained above, are primarily due to the use of different methods of valuing both housing and businesses.

Overall, in estimating total assets (financial and non-financial combined), we are pleasantly surprised that HILDA appears not to be substantially under the mark. While gratifying, this is something of a puzzle given the more-or-less unavoidable underrepresentation of the very wealthy. The wealthiest household included in the HILDA sample, for example, had a reported net worth of $22 million, well below the levels recorded for individuals listed in the Business Review Weekly's list of Australia's 200 wealthiest people (Business Review Weekly 2004).

Debts

The final comparisons we discuss in this chapter relate to debts. Comparability among the three data sets here is much greater, so we can reasonably just review total debt estimates rather than discuss each component separately.

HILDA: $517 billion

ABS: $631 billion

RBA: same data as ABS

Here the HILDA estimate is only 82 per cent of the ABS figure. In retrospect, we suspect the HILDA questionnaire may not have included enough questions on separate types of debt. As noted above in Table 1, HILDA split household debt into five categories: housing debt, business debt, student debt (HECS), credit card debt and 'other' personal debt. It might have been preferable to ask specifically about overdrafts (excluding housing), vehicle debt, hire purchase, gambling debts and so on (see Juster, Smith & Stafford 1999). Even so, there may be some irreducible tendency for respondents to underreport debt, partly for social desirability reasons. We also believe that, relative to official sources, credit card debt will be understated in HILDA. Those respondents who said they routinely paid up in the first month and so incurred no interest charges were recorded as having no credit card debt. By contrast, the official sources record credit card liabilities owed by the nation's households at one moment in time.

[ Return to Top   Return to Section ]

3. Findings

3.1 Introduction

All results are weighted to adjust for differences between sample and population characteristics. The results can thus be treated as population estimates, or as being weighted up from a sample size of 7,245 households to a population size of 7,540,411 households. Because this is a large sample the confidence intervals for most estimates given in this report are within plus or minus 3.5 per cent at a confidence level of 95 per cent. Of course, where estimates of the wealth of smaller subsets of the population are given, they are less reliable. Estimates which would be too unreliable for most practical purposes because they have a standard error of more than 50 per cent of the estimate have been marked 'nr' in the tables that follow.

3.2 The structure/composition of wealth

Issue 1: What are the main components of wealth and debt? What are the relative shares of financial and non-financial (particularly housing) assets?

Table 2 gives an overview of the wealth of Australian households in the last quarter of 2002.4 It gives mean and median assets and debts, and hence net worth, and also the percentage contribution which each type of asset and debt makes to total holdings.5

It should be noted that the medians reported in Table 2 (but not later tables) are somewhat unusual. The aim is to describe the wealth of the typical Australian household. So we report the median assets and debts of households in the 50th (median) percentile of net worth. In other words, we take households with overall wealth (net worth) that is 'typical' and then show their asset and debt levels. Because the distribution of wealth is highly skewed, medians give a better idea of the typical household's wealth than do means.

In the last quarter of 2002, the average household had an estimated net worth of approximately $404,800 ($473,300 of assets and $68,500 of debts). However, these mean estimates are skewed upwards by inclusion of the rich. The median household had assets of only about $288,000 and a net worth of about $218,600.

As is well known, Australians' asset portfolios are dominated by housing. Housing and other property constitute more than 50 per cent of all household assets and more than 60 per cent of the assets of the median household. Over two-thirds of households—67.8 per cent—owned or were buying their own home. Quite a high proportion of Australian households—16.7 per cent—had invested in other property as well (a holiday home or investment property).

The second largest asset of most households is superannuation, which has become much more widely held and somewhat more equally distributed in the last 15 years (Kelly 2001). Even so, the median household holds superannuation worth only about $17,000. Other holdings of considerable value to some households are business assets and equity investments (for example, shares, managed funds, listed property trusts). The median household holds no equities and does not own a business. However, the 41 per cent of households that do own equities average about $70,000 worth (median=$15,000), and the average value of businesses (owned by 13 per cent of households) was about $291,000 (median=$80,000). It should be noted, however, that equity investments are understated here since, in order to avoid double counting, HILDA respondents were asked not to include superannuation in their calculation of equity holdings; and of course some superannuation was held in equities. Moving then towards the bottom of the list of assets, the median household had vehicles worth about $12,000 and $4,700 in the bank.

Household debt is mainly mortgages. The average property debt is about $51,000. Most households have little or zero in other forms of debt.

Table 2: Overview: assets, debts and net worth of households in 2002, population weighted estimates
n=7,245 households Mean
($'000)
Median
($'000)
% total assets
or debts
% households
holding
assets/debts
Overall assets and debts        

Total assets

473.3 288.0    

Total debts

68.5 10.0    

Net worth (assets minus debts)

404.8 218.6    
Assets in order of value        

Housing and other property

255.0 180.0 53.9 71.0(a)

Pensions/superannuation

75.2 17.0 15.9 77.0

Businesses and farms

44.4 0 9.4 13.1

Equity investments (shares, managed funds)

31.3 0 6.6 41.4

Bank accounts

21.4 4.7 4.5 97.3

Cars and other vehicles

19.0 12.0 4.0 87.9

Other assets(b) (c)

27.9 0 5.9 47.4
      (100.0)  

Non-financial assets

315.4 204.5 66.6 93.6

Financial assets(c)

157.9 49.5 33.4 99.3
  (473.3)   (100.0)  
Debts in order of value        

Housing and other property

51.4 0 75.0 38.7

Businesses and farms

6.8 0 9.9 5.2

HECS (student) debt

1.3 0 1.9 12.7

Credit (and similar) cards

1.1 0 1.6 39.5

Other debts (cars, hire purchase etc.)(c)

7.9 0 11.5 36.7
  (68.5)   (100.0)  

(a) 71.0 per cent of households owned property; 67.7 per cent owned the home they were living in. (b) Other assets include cash investments, trust funds, the cash-in value of life insurance and collectibles. (c) Small adjustments have been made to these three items in order for totals to balance. The reason for what would otherwise be small discrepancies is that the imputations of wealth undertaken by the RBA omitted five components asked on the Person Questionnaire: bank accounts, superannuation, credit card debt, HECS debt and other personal debt. The authors imputed these items but did not constrain the imputation to force the total of all components to equal the previously imputed total asset and total debt values. Our intention is to revise the imputation to address this problem. Note: All results are weighted up to population size. The national population of households was 7,540,411.

Overall, non-financial assets dwarf financial assets. Most households lack liquidity. They have little cash and little that they can easily cash up if normal sources of market income are temporarily or permanently cut off or if emergency expenditures are required. This means they must rely primarily on pension and benefit entitlements. This is especially clear when one remembers that, until one retires, superannuation is not available and so, while classified as a financial asset, it is not liquid. Some could perhaps obtain loans in an emergency, although, as noted above, it is difficult for low-income households to obtain credit at reasonable interest rates.

Issue 2: Has the level and composition of household wealth changed in recent years? If so, to what extent have increases been due to possibly transient increases in housing values?

To date, the HILDA survey has only asked about wealth once—in 2002. So from HILDA we cannot directly answer questions about trends in wealth. However, all published estimates, both from the Australian Government (ABS, cat. nos 5232.0 & 5204.0) and from academic sources (Kelly 2001, 2003) show total household assets and net worth increasing faster than the rate of inflation in recent years. For most of the 1990s, stock market and property prices grew strongly. However, stock market values grew more rapidly than property prices. Then in the period 2000 to 2003, world stock market values fell sharply, although the Australian market fell much less, and property prices rose rapidly in all capital cities. So in this latter period the rise in non-financial asset values far exceeded the rise in financial assets (which barely kept up with inflation).

Another key point is that Australian households are changing their asset portfolios. Property still dominates, but financial assets in the form of shares and superannuation are becoming increasingly significant (Kelly 2001, 2003). (Trends in superannuation are discussed in some detail under Issue 4.1.)

The question as to whether recent increases in household wealth are likely to prove transient because they are due to housing price increases is thus hard to answer. On the one hand, it may well be that in most future years either the stock market or the housing market will do well and that most households will be positioned to gain some advantage from either development. On the other hand, it remains true that most households still stand to prosper more from housing market gains than stock market gains.

3.3 The distribution of wealth

Issue 3a: What is the overall distribution of wealth?

In Australia, as in other Western countries, wealth is much more unequally distributed than income (Atkinson, Rainwater & Smeeding 1995). The most commonly used measure of distribution is the Gini coefficient which ranges between 1 (all wealth is held by one household) and 0 (every household has the same wealth). In 2002 the Gini coefficient of household net worth was 0.624. This can be compared with a Gini coefficient of 0.422 for household gross incomes and 0.382 for household disposable incomes (see Table 3).

It should be noted that only moderate correlations are found between wealth and income. In HILDA, the correlation of household net worth with household gross income was 0.35, and with net income it was 0.34.6 Such correlations may seem surprisingly low, but similar results are found in other Western countries (Klevmarken, Lupton & Stafford 2003). If analysis is restricted to households headed by prime-age men and women (aged 25 to 55), the correlations are somewhat higher at 0.40 for gross income and 0.39 for net income.

The findings that wealth and income are not highly correlated and that wealth inequality is greater than income inequality are both primarily due to the greater dependence of wealth on age, or rather on saving as one ages. Wealth also depends somewhat on inheritance although, contrary to widespread impressions, most wealthy people are 'self-made' rather than being beneficiaries of large inheritances—see, for example, the list of Australia's 200 wealthiest people (Business Review Weekly 2004). So wealth accumulates primarily through both voluntary saving and compulsory superannuation, and these savings compound as people age.7 Income also increases with age but the gradient is not as steep as wealth's compounding gradient.

Table 3 presents summary data on the distribution of assets, debts, net worth and their main components. Also included for comparison are measures of income distribution.

Table 3: The distribution of wealth and income, Gini coefficients
Assets/debts/income Gini
Total assets 0.590
Total debts 0.757
Net worth 0.624
Gross income 0.422
Net income 0.382
Specific types of assets/debts Gini
Property assets 0.588
Superannuation 0.751
Vehicles 0.563
Bank accounts 0.772
Property debt 0.791

Table 3 gives Gini coefficients for the types of assets and debts held by most Australian households. It can be seen that, relatively speaking, housing assets and vehicle values are somewhat less unequally distributed than total assets, while superannuation and bank savings are more unequally distributed. Debts show even greater dispersion than assets—the Gini of total household liabilities being 0.757, with the Gini for property debt at 0.791. In explaining these differences, a key point is that assets and debts are positively correlated (r=0.45). The reason for this initially surprising correlation is that the more you own, the more you can borrow. Well-off households can readily obtain loans at reasonable rates of interest, whereas poor households cannot. A similar point applies to savings. Households with high incomes find it much easier to save, and once a reasonable level of assets has accumulated, they are more likely to seek higher risk–return ratios and will therefore, in general, tend to accumulate savings more rapidly than less well-off households. Hence the finding in Table 3 that bank savings and superannuation savings are highly concentrated.

The distribution of wealth among older households (reference person aged 65 and over) is a little less unequal than it is in the total population. The Gini of net worth for this group is 0.563 and for assets it is 0.560.

Another method of summarising the distribution of wealth is to show the shares owned by various quantiles. Table 4 gives the shares of total net worth owned by each decile and by the wealthiest 5 per cent.

Table 4: Shares of total wealth (net worth) by deciles
n=7,245 households Share (%) Median ($'000)
Wealthiest decile 44.9 1,394.3
(wealthiest 5%) (31.0) (2,511.8)
9th decile 18.2 727.2
8th decile 12.4 498.9
7th decile 9.0 364.7
6th decile 6.5 262.1
5th decile 4.5 181.8
4th decile 2.8 113.6
3rd decile 1.3 54.5
2nd decile 0.4 14.0
Least wealthy decile negative 0

The HILDA data indicate that, in 2002, the wealthiest decile owned 44.9 per cent of total household wealth (median holdings=$1,394,400) and the wealthiest 5 per cent owned 31.0 per cent (median=$2,511,800). For reasons outlined in Section 2, it is likely that we somewhat underestimated the assets and national share of the richest households.

Surprisingly, then, our estimates for the 'top end of town' are in fact a little higher than recent imputed estimates derived mainly from ABS national aggregate measures. Kelly (2001) estimated that the top 5 per cent held 30 per cent of net worth in 1998 and Northwood, Rawnsley and Chen (2002) estimated that, in 2000, the top decile held 43 per cent, with the top quintile holding 61 per cent. If it is true that the share of the top end has increased in the few years since these imputed estimates were made, it is likely to be due to rapid house price increases in all the main capital cities.

HILDA indicates that, by 2002, 8.8 per cent of households had a net worth over $1 million, with 11.2 per cent holding assets over the million dollars mark. So, in a sense, about 1.8 million people living in 638,000 Australian households could be described as millionaires.

At the lower end of the distribution, the poorest decile of households on average have debts exceeding their assets (negative net worth), with median net worth being just $24. In all, the least wealthy half of the population owns only 9 per cent of total household net worth.

Issue 3b: How are wealth and indebtedness distributed among age cohorts? How wide is the gap between the most and least wealthy cohorts? How is wealth distributed within cohorts?

In this sub-section we review cohort differences in household wealth. As is already clear, wealth is heavily affected by age. From a public policy standpoint the key issue is whether cohorts approaching retirement and those recently retired have adequate assets to enjoy a reasonable lifestyle after they finish paid work.

In Table 5 and subsequent tables, 'couple households' are classified by the age of the male partner.8 In lone-parent households the 'reference person' is the lone parent and in single person households it is that person.9 Similarly to many Australian Government publications, we have divided households into those with reference persons in the 15 to 24 years cohort, then 25 to 34, 35 to 44, 45 to 54 and so on.

Table 5 gives an overview of the differences between and within cohorts by focusing just on net worth. It shows the mean (average) net worth of each cohort and then the net worth at the midpoint of each quintile—that is, at the 10th, 30th, median, 70th and 90th percentiles of the distribution.

Table 5: Age cohorts, net worth of households at the cohort mean and at the 10th, 30th, median, 70th and 90th percentiles
Household reference person's age

Net worth

Mean ($'000) 10th percentile ($'000) 30th percentile ($'000) Median ($'000) 70th percentile ($'000) 90th percentile ($'000)
15–24 28.3 –8.5 0.2 5.0 17.0 89.0
25–34 162.6 0.8 24.3 74.6 159.7 385.0
35–44 340.9 7.0 83.9 204.8 381.0 727.7
45–54 521.3 29.5 183.5 361.7 580.0 1,130.1
55–64 671.8 17.1 216.0 422.1 741.5 1,508.8
65–74 530.3 19.9 181.0 318.0 538.0 1,127.0
75+ 348.8 15.3 138.0 244.5 361.3 768.0
All 404.8 4.2 83.0 218.6 428.0 934.2

Two contrasting results show through clearly in this table. The first is the strong dependence of wealth on age or, really, time spent saving and investing. The second which, while not contradictory, points in a different direction—that is, even within age cohorts there are great disparities in wealth.

Let us examine the evidence for each cohort. The poorest cohort is the youngest one (reference person aged 15 to 24 years) with a median net worth of just $5,000. At the 10th percentile (the middle of the poorest quintile), households have negative net worth: their debts exceed their assets by $8,500. At the 90th percentile net worth is $89,000. Contrast this quintile's situation with that of the wealthiest cohort, namely the quintile of households whose reference person is aged 55 to 64 years. Just under 60 per cent of these households are still moving towards, and are saving for, retirement (see Table 7). The rest have quite recently retired and have (presumably) not yet run down their savings by much. In this quintile median net worth stands at $422,100, but even here there are large disparities, with net worth being only $17,100 at the 10th percentile and $1,508,800 at the 90th percentile. The oldest cohort (reference person aged 75 and over) is from a generation which was always less well off than younger generations. Furthermore, after retirement people usually run down their savings (although many may be determined to leave substantial bequests to their partners and/or children) with the result that this cohort has a median net worth of $244,500. In this group also disparities are vast, with a net worth of $15,300 at the 10th percentile compared with $768,000 at the 90th percentile.

3.4 Retirement issues

Issue 4a: Retirement—focusing on the older age cohorts (those aged 45 years and over), what is their capacity for self-funding in retirement?

We now focus more closely on retirement issues by looking in detail at the financial assets and incomes of the cohorts that are approaching retirement or have already retired. As background it is useful to know that at present about 68 per cent of retirees receive a full means-tested Age Pension, with another 15 per cent receiving a part pension.

What level of wealth—what investable sum—do people require at retirement to generate an adequate income for the rest of their life? This much-debated question has led to a great deal of research within and outside government (ASFA 1999, 2004a, 2004b; Kelly 2001, 2003). We do not pretend to be experts in this highly specialised field (a field which requires actuarial skills), but in order to address retirement issues, we need to make some assumptions and estimates. The following tables are directly based on HILDA data, but it should be understood that the commentary is tentative and not an expert view.

A frequently used rule of thumb in the superannuation industry is that 60 per cent of pre-retirement gross income is considered adequate. However, this rule is not as straightforward as it seems. Most people do not earn their maximum gross income directly before retirement. Generally, a person's income peaks in their late 40s and early 50s and then tapers down before retirement. In this report, we interpret the rule of thumb to mean that people should aim for target incomes during retirement which are 60 per cent of the gross income earned by the couple (in a couple household) or the single person (in a one-person household) during their peak earning years.

Other rules of thumb relate to the amount of financial assets (that is, not including the family home or other tangible assets) one needs to invest to generate a target level of income. For example, the Australian Superannuation Funds Association (ASFA) suggests that someone who retires at 55 will need an invested sum that is 17 times the target level of income. Retirement at 60 lowers this multiplier to 15 times the target level of income and, at 65, the multiplier is 13.10 It should be noted that ASFA estimates take account of entitlement to a full or part pension in the case of individuals and couples whose non-pension incomes and assets do not exceed the prescribed limits.

Some further assumptions must be made to estimate future retirement income. In the case of people still working, we assume that the financial assets which they reported to HILDA in 2002 will grow at 3 per cent per year in real terms (that is, after allowing for inflation) up to retirement. We also assume that funds invested under the Superannuation Guarantee (9 per cent of income) between 2002 and date of retirement will grow by 3 per cent. These assumptions, taken together, make it possible to estimate the investable sums likely to be available at retirement.

Estimating household income after retirement is much more problematic. The ASFA rule of thumb given above is a rough guide. However, models developed by the Treasury, by life insurance companies, by ASFA and by financial advisers yield substantially different results. It is reasonable to assume that the Australian Government will continue to adjust the pension so it remains at 25 per cent of average weekly earnings. But estimating likely future earnings from households' own savings in superannuation funds and other sources is fraught with difficulty. Returns depend on the risk which households (or their financial advisers) are willing to take and on future returns for different types of investment (shares, property, bonds and cash deposits). The risk issue is crucial. Most available models appear to make conservative assumptions about the risk profiles of retired people, presumably on the grounds that they need a steady income and therefore will not want to make investments which are high risk/high return. The 3 per cent rate of return which we use is moderately conservative. Plainly, however, a retiree with a high level of financial assets might sensibly choose to take higher risks in expectation of higher returns.

In making their investment decisions at retirement, we assume that people take account of their public pension entitlements. So in the calculations given below, it is assumed that financial assets are used to purchase an income stream (for example, a complying annuity) which preserves the maximum entitlement to the Age Pension consistent with the annual income stream they are receiving. In September 2002, when the HILDA survey was conducted, the full single Age Pension stood at $11,164 per year and the couple pension at $18,637. However, the pension reduces on a sliding scale once certain income test and asset test limits are reached. In September 2002 the income limit at which the pension for a home owner couple started to reduce was $5,304. The pension then cut out completely at an income of $52,273.

A further issue is that retired households' actual incomes and standard of living partly depend on whether people are willing to use up all their savings before death. Models used by government—and the complying annuity method of investment used here—assume retirees are willing to exhaust savings even though it is known that many people have a 'bequest motive' and want to leave assets to their heirs.

To avoid making this report excessively complicated, we review evidence and estimates about the financial assets and incomes of four pre-retirement cohorts and four post-retirement cohorts. The pre-retirement cohorts live in households in which the reference persons in 2002 were aged 45 to 49 years, 50 to 54, 55 to 59 and 60 to 64.11 The post-retirement cohorts are those with reference persons aged 60 to 64 years, 65 to 69, 70 to 74 and 75 and over. Projections of future post-retirement incomes for those still working are made on the basis of two assumed ages of retirement—60 and 65. It could be argued that, for most people, 60 is the more likely age. At present in Australia about 53 per cent are retired by their 61st birthday (49 per cent of men and 57 per cent of women), and about two-thirds of HILDA respondents say they wish to retire by 60 (Melbourne Institute 2003).

As well as dividing the population into age cohorts, we also distinguish between couple households and households headed by non-partnered persons. Then, within these categories, we distinguish between home owners and renters. These distinctions are almost always made in analyses of retirement issues, superannuation and pensions. The Australian Government sets different pension levels and income test limits (affecting pension entitlements) for couples and singles, and different asset test limits for home owners and renters.12 The rationale is that couples do not need twice the income of singles in order to have the same standard of living (partly because they share a home), and that people who have paid off their home are, other things equal, better off than renters.

Just over 70 per cent of individuals approaching retirement age are living in couple households which own or have nearly paid off their home. Quantitatively they are the largest group, so we will focus most of the analysis on them. We will also assume the home is fully paid off by retirement and that there are no other debts. However, some attention will also be paid to the three worse-off groups: couples who are not home owners (15.5 per cent in September 2002); singles who are home owners (about 7 per cent); and singles who are renters (also about 7 per cent).

Couple households who are home owners

Let us begin by examining the financial assets and incomes of couple home owners who in 2002 were already retired. We focus on cohorts in which the household reference person is aged either 60 to 64 years, 65 to 69, 70 to 74, or 75 and over. These data give us a picture of the current situation of retired people and provide yardsticks of comparison for assessing the future prospects of cohorts still approaching retirement. All estimates are in constant prices in September 2002 dollars.

The aim in Table 6 is to provide an assessment of the living standards of 'typical' households, so analysis is restricted to households falling between the 25th and 75th percentiles of 'couple combined income' for the cohort in question.13 By excluding the highest and lowest income quartiles, we obtain mean (average) and median results that are quite close to each other (except for assets) and that give us a clear initial picture of typical households.

Table 6: Financial assets and incomes of retired couple home owners, cohorts aged 60–64, 65–69, 70–74 and 75+, population weighted results
n=244 households

Reference person
60–64 retired

Reference person
65–69 retired

Reference person
70–74 retired

Reference person
75+ retired

Mean
($'000)
Median
($'000)
Mean
($'000)
Median
($'000)
Mean
($'000)
Median
($'000)
Mean
($'000)
Median
($'000)
Financial assets 215 144 256 163 127 86 100 42
Household gross income 30 32 30 28 24 23 23 22
Household net income 28 29 28 28 23 23 23 22
Household asset income 3 1 5 1 4 2 3 1
Household private pension 11 5 8 3 4 0 1 0
Household public transfers (mainly Age Pension) 10 11 13 17 15 18 18 18
Male partner wage 3 0 1 0 0 0 0 0
Female partner wage 3 0 2 0 0 0 0 0

Note: Amounts are given in $000s to the nearest $1,000. Income components do not exactly sum to total household gross income, due both to rounding and to omission of minor components of income, including inter-household transfers.

Table 6 shows a pattern we will see repeated throughout the analysis. The two youngest of these four cohorts (aged 60 to 64 and 65 to 69) are best off, with the highest levels of household gross and net incomes. They are better off than the middle cohort (aged 70 to 74) which, in turn, is a little better off than the oldest cohort (aged 75 and over). The main reason for this pattern is that each successive cohort receives somewhat higher real incomes during its working lifetime and so tends to save more. This is, as a general trend, indicated by higher levels of financial assets (row 1 in Table 6). However, the 60 to 64 retired cohort is an exception to this trend and appears to be comprised differentially of relatively low net worth households who have nevertheless retired or been pressured into retirement early.14 A second reason for the general trend towards rising household assets is that the Superannuation Guarantee, which began in 1993, is gradually increasing savings and hence the investable sums available at retirement.

It is not possible to calculate what percentage of their pre-retirement (peak) gross earnings these cohorts are now receiving in retirement, because HILDA did not ask respondents about past earnings. However, it seems unlikely that any of them are receiving much more than 50 per cent. The 65 to 69 cohort is a case in point. They average $30,000 of household gross income, of which the largest segment comes from public pensions, with smaller amounts coming from private pensions and asset income. Based on the observed pattern of earnings for younger cohorts (given in later tables), it is unlikely that at their peak these couples were earning substantially less than $60,000.

We now consider households in which the reference person is not yet retired. The aim is to estimate their likely investable savings at retirement and also, within a range, their annual income during retirement. Two retirement ages are allowed for—retirement at 60 (except for the 60 to 64 cohort which is already past that age) and retirement at 65. In Table 7, column 2 refers to the financial assets reported in the HILDA survey in 2002 and column 3 gives the couple's combined labour income on which 9 per cent superannuation contributions are based from 2002 until retirement. Column 4 estimates accumulated financial assets by retirement at age 60, assuming a 3 per cent real rate of return between 2002 and date of retirement. Column 5 then estimates the annual income which these assets would generate from retirement until death at age 82 (the assumed average age of death), again assuming a 3 per cent rate of return. Columns 6 and 7 are the same as columns 4 and 5, but with an assumed retirement age of 65. For all cohorts, retirement income consists partly of a private pension income, generated by a lifetime annuity or similar investment, and partly of public transfers (the Age Pension for women aged 60 and over and men aged 65 and over). In other words, the estimates of future income in columns 5 and 7 take account of rules about eligibility for a full or part pension.15

Table 7: Projected financial assets and income, retirement at 60 or 65, population weighted results
1 2 3 4 5 6 7
 

Median ($'000)(a)

Age of household
reference person (n=538 households)
Financial
assets
2002
Couple's
labour
income
2002
Projected
financial
assets
— retired
at 60
Projected
income
— retired
at 60(b)
Projected
financial
assets
— retired
at 65
Projected
income
— retired
at 65(b)
45–49
(15–20 years to retirement)
111 70 261 30–32 336 35–38
50–54
(10–15 years to retirement)
203 67 311 31–34 392 37–40
55–59
(5–10 years to retirement)
211 53 245 29–31 310 33–36
60–64
(0–5 years to retirement)
202 35 na na 230 30–32

(a) Amounts are given in $'000 to the nearest $1,000. Income components do not exactly sum to total household gross income, due both to rounding and to omission of minor components of income, including inter-household transfers. (b) Annual income from age 60 or 65 until death at age 82 (September 2002 dollars).

Table 7 again suggests that, in general, the younger cohorts are likely to do better in retirement than the older ones. An exception appears to be the youngest cohort (reference person aged 45 to 49). However, the estimates for this group are almost certainly too low. The problem is their low level of financial assets reported in 2002. In practice, households do not save a constant proportion of income throughout the last 20 to 25 years of working life. As retirement gets closer and children leave home, savings rates accelerate. The 45 to 49 cohort, shown in Table 7, had clearly not yet accelerated its savings rate, but it had a high median income in 2002 ($70,000), and so will probably eventually save enough to be as well as, or better off than, the cohort preceding it.

Which of these cohorts meets the target of 60 per cent of pre-retirement (peak earning period) income? It is hard to be sure because of difficulty in estimating peak earnings. However, the case of the 50 to 54 cohort gives us some basis for extrapolating to other cohorts. The 50 to 54 cohort was presumably at or near its couple-earning peak in 2002 and had a median income of $65,000 at that date. The estimates in Table 7 indicate that, at the median, this cohort is likely to receive an income of just under 50 per cent of this figure if retirement occurs at 60, and just under 60 per cent if retirement occurs at 65. If we were to guesstimate that the peak earning figure for the 55 to 59 cohort was, say, $60,000 (rather than the $55,000 reported in 2002 when they were presumably past their peak), then they are in much the same position; that is, they are likely to receive around 50 per cent of peak earnings if they retire at 60 and around 58 per cent if they retire at 65. Guesstimating for the other two cohorts—the youngest and the oldest—is more difficult. However, some reasons have already been given for thinking that the youngest cohort will eventually be at least as well off as the cohort preceding it. Perhaps in the case of the oldest cohort, a reasonable guess would be that their peak couple income was $50,000 or a little more. If so, retired at 65, they too are likely to receive close to 60 per cent of peak income.

Couples who rent and households headed by non-partnered persons

The 30 per cent of the population who are not in home owner couple households are much worse off both prior to and after retirement. The groups analysed here are couple households who are renters, and then two types of households headed by non-partnered persons: (i) those who own a home and (ii) those who rent. Because these are much smaller population groups than home owning couple households, and are therefore represented in smaller numbers in HILDA, analysis is confined to more broadly defined cohorts than were considered earlier. The focus is on retired households with a reference person aged 65 to 74 and non-retired households with a reference person aged 50 to 59. In the case of households headed by non-partnered persons, we distinguish between male and female heads.

Table 8: Financial assets and incomes of couple renter households and households headed by non-partnered people, population weighted results
n=559 households

Couples who rent

Non-partnered owners

Non-partnered renters

Retired
65–74

Not retired
50–59

Retired
65–74

Not retired
50–59

Retired
65–74

Not retired
50–59

Men Women Men Women Men Women Men Women
Financial assets
($'000)(a)
3 46 58 30 55 56 2 5 33 14
Household gross
income – retired
($'000)
20 13 14 11 12
Labour income
– not retired(b)
($'000)
35 31 30 22 22

(a) Amounts are given in $000s to the nearest $1,000. Income components do not exactly sum to total household gross income, due both to rounding and to omission of minor components of income, including inter-household transfers. (b) Couple combined labour income for couple households and individual labour income for households headed by non-partnered people.

Despite the small subsample sizes, the overall pattern is clear. All these households have little in the way of financial assets. Those who have retired (reference person aged 65 to 74) are living off the Age Pension, supplemented by a small amount of income derived from personal savings. Those in the 50 to 59 cohort and not retired have very low levels of financial assets and 2002 wages well below average. So they too will almost certainly need to rely almost entirely on the Age Pension when they retire.

Issue 4b: Is there a relationship between intended age of retirement and levels of wealth and debt?

It seems a reasonable hypothesis that people who want to retire at a relatively young age will save and accumulate more assets at an earlier age than people who want to retire at a later age. The equation below relates to household reference persons and the dependent variable is 'age of intended retirement' (a question put to all HILDA respondents aged 45 and over). The explanatory variables of main interest are the household's asset holdings and the reference person's self-reported savings behaviour. Respondents were divided into three groups according to their savings behaviour: regular savers, irregular savers and non-savers. Household income was also included in the equation because it was thought that higher income households could afford to retire earlier. Then additional variables were included essentially as 'controls'. These were the reference person's age and square of age (to capture non-linear effects). (Some further controls were tried, including household size and household type—couple with dependants, couple with no dependants, single parent household and so on—but these variables proved non-significant and are not shown in Equation 1.)

Equation 1:

Intended retirement age = –3.87 – 0.36 Assets (ln)*** – 0.86 Regular Saver* – 0.92 Irregular Saver** + 0.52 Age*** – 0.05 Age 2***– 0.37 Household Gross Income (ln)*

Note: R2=2.4% (n=4,783)
***significant at 0.001 level; **significant at 0.01 level; *significant at 0.05 level.
The reference category for the saving variable was 'non-saver'.

The results can be interpreted as follows. First, household reference persons with higher levels of assets intended to retire at a younger age. The relationship (b= –0.36) between asset holdings and age of intended retirement was significant at the 0.001 level. However, despite statistical significance, the size of the effect was quite small.16 There was also an interesting and statistically significant relationship between reported savings behaviour and retirement intentions. Both those who reported 'regular saving' and those who reported 'irregular saving' intended to take earlier retirement than those who were currently not saving any money. Not surprisingly perhaps, reference persons in higher income households expected to retire somewhat earlier, although this relationship was only significant at the 0.05 level.

Age and age-squared were included in the equation mainly as controls, so that we can say that the results of main interest relating to assets and savings behaviour hold true at any given age. However, there is also a result of substance here: the older the HILDA respondents were at time of interview, the later was their intended retirement age (b=0.52). This was true up to just over age 50. After that, intended retirement age fell (b= –0.05 for age2).

The results in this section should be treated with caution. They only relate to the relationship between stated intentions about retirement on the one hand, and asset holdings, incomes and savings behaviour on the other. Actual behaviour may or may not turn out to be different from stated intentions. It should be noted that in the 'mixed' 60 to 64 years cohort—a cohort about evenly divided between retired and non-retired households—the households which had in fact already retired had lower—not higher—asset levels than those still working. It is hard to separate cause and effect here: doubtless one reason non-retired households had more assets was precisely because they were still working. Even so, until further evidence of actual behaviour emerges, it would not be sensible to place much weight on the apparent relationship between financial behaviour and retirement intentions.

3.5 Vulnerable groups in society

Issue 5: How do vulnerable and 'at risk' groups fare in regard to wealth? In particular, what are the assets and debts of income support recipients? What is the wealth situation of lone parents?

We now describe the assets and debts of two 'at risk' groups: households headed by an income support recipient and those headed by lone parents. The term 'income support recipients' refers to people receiving FaCSIA/Centrelink payments, but not if they only receive Family Tax Benefit and/or Child Care Benefits. Table 9 focuses on households with heads who are not retired and so are not receiving the Age Pension.17>

Table 9: Assets and debts of income support recipients, household reference person receiving income support and not retired
n=1,085 households Mean
($'000)
Median
($'000)
%
Overall assets and debts      

Total assets

143.7 30.0  

Total debts

25.2 4.0  

Net worth (assets minus debts)

118.5 18.7  
Assets      

Housing and other property

87.4 0 60.8

Pensions/superannuation

21.3 1.5 14.8

Businesses and farms

4.7nr 0 3.3

Equity investments (shares, managed funds)

7.5 0 5.2

Bank accounts

6.6 0.6 4.6

Cars and other vehicles

8.6 3.5 6.0

Other assets(a)

7.6 0 5.3
  (143.7)   (100.0)

Non-financial assets

104.0 10.0 72.4

Financial assets

39.7 4.3 27.6
  (143.7)   (100.0)
Debts      

Housing and other property

18.4 0 73.0

Businesses and farms

1.0nr 0 4.0

HECS (student) debt

1.9 0 7.5

Credit (and similar) cards

0.7 0 2.8

Other debts (cars, hire purchase etc.)

3.2 0 12.7
  (25.2)   (100.0)

(a) Other assets include cash investments, trust funds, the cash-in value of life insurance, and collectibles. Note: nr Not reliable—standard error over half the estimate.

Table 9 makes it clear that income support recipients are short of wealth in all respects. Their average net worth is $118,500 with a median of just $18,700. This compares with an Australian average of about $405,000 (median=$218,600), as reported in Table 2, and an average for households headed by non-income support recipients (not retired) of $397,200 (median=$231,500). Despite similar age profiles, the value held in all asset categories is much lower than for non-retired households in which the reference person does not receive income support. Superannuation holdings are particularly low at a mean of $21,300 and a median of just $1,500. This compared with a mean of $88,600 for households where the reference person was not receiving an income support payment. It appears likely that most current income support households are headed towards reliance on the Age Pension when retirement comes. However, it should also be noted that debt levels in this group are quite low, although this is likely due in part to difficulty in accessing credit without an asset base.

The difficulties of income support recipients, especially those close to retirement, are compounded by low rates of home ownership. Only 38.4 per cent were owner occupiers compared to 69.2 per cent of non-retired households where the reference person did not receive income support.

Table 10 shows the assets and debts of households headed by lone parents—a particularly vulnerable group among whom nearly two-thirds (65.8 per cent) were income support recipients.

Table 10: The assets and debts of lone-parent households
n=528 households Mean
($'000)
Median
($'000)
%
assets/debts
Overall assets and debts      

Total assets

200.9 65.5  

Total debts

36.5 4.5  

Net worth (assets minus debts)

164.4 47.8  
Assets      

Housing and other property

123.0 0 74.8

Pensions/superannuation

27.3 3.0 16.6

Businesses and farms

6.8nr 0 4.1

Equity investments (shares, managed funds)

19.5 0 11.9

Bank accounts

8.2 0.9 5.0

Cars and other vehicles

8.9 5.0 5.4

Other assets(a)

7.2 0 4.6
  (200.9)   (100.0)

Non-financial assets

141.3 26.0 70.3

Financial assets

59.6 5.5 29.7
  (200.9)   (100.0)
Debts      

Housing and other property

28.5 0 78.1

Businesses and farms

1.1nr 0 3.0

HECS (student) debt

0.9 0 2.5

Credit (and similar) cards

1.0 0 2.7

Other debts (cars, hire purchase etc.)

5.0 0 13.7
  (36.5)   (100.0)

(a) Other assets include cash investments, trust funds, the cash-in value of life insurance, and collectibles. Note: nr Not reliable—standard error over half the estimate.

Lone-parent households have considerably less wealth than most other non-retirement age households. Lone parents are just a few years younger on average than the comparison group, being more concentrated in the cohorts aged 25 to 34 and 35 to 44 and less in the cohort aged 45 to 54. Even so, and allowing for more years to accumulate assets, it is clear that their median net worth of $47,800 lagged well behind the net worth of other non-retired households (median=$231,500). On the other hand, lone parents had higher typical net worth than the entire group of income support recipient households (Table 9), but this was entirely due to a third of them not being income support recipients. If we restrict the comparison to lone parents who are income support recipients, it becomes clear that they are worse off than the rest of this group. Their average net worth amounts to $69,100 (median=$8,000), only 33.1 per cent are owner occupiers and their superannuation amounts to $8,200 (median=$1,000).

3.6 Determinants of wealth

Issue 6: What are the main factors (demographic, educational, income related, and so on) which determine levels of wealth and debt?

Previous sections of this report have examined in some detail the assets and debts of sections of the Australian community. We now try to provide a somewhat fuller statistical account of differences in wealth among Australian households. In these analyses the dependent (outcome) variable is net worth and the explanatory variables are a range of demographic, educational, health, income and attitudinal variables which are hypothesised to relate to wealth.

Three sets of regression results are given: one for households headed by prime working age people (aged 25 to 54), one for the traditional pre-retirement decade (aged 55 to 64) and one for households headed by retirees aged 65 and over.

Variables were entered into the regressions in three steps which were seen as following the time sequence of most people's lives. The steps were:

  • Step 1—Characteristics one is born with: gender of the household reference person (f=1, m=0), occupational status of the reference person's parents; reference person is Australian born, born in another English speaking country, born in a non-English speaking country (reference group). The Australian National University occupational status scale was used to classify parental status: the father's occupational status was taken if available, the mother's if not. We hypothesised that households would be wealthier if the reference person was male and came from a high status background. We also hypothesised that households headed by Australian born and people from other English speaking countries would be wealthier than households headed by immigrants from non-Anglo backgrounds.
  • Step 2—Education: university degree, trade qualification, completed Year 12 (reference group), did not complete Year 12. We hypothesised that wealth would co-vary with the reference person's level of education.
  • Step 3—Household type, health, hours worked and income, and attitudinal variables: couple with no dependants, couple with dependent children or dependent students, lone parent with dependants, one-person household (reference group). The hypothesis was that couple households would be wealthier than one-person households and that lone-parent households would be worst off. The SF-36 scales of physical and mental health were used, both scored 0–100.18 We expected that households headed by people in good physical and mental health would be wealthier. The natural log of the reference person's average weekly working hours and the household's gross annual income were included. We expected that both variables would be positively related to wealth. Dummy variables measuring self-reported saving behaviour and attitudes to financial risk (risk aversion) were also entered. We expected that self-reported saving behaviour would be associated with greater wealth and that risk aversion would be associated with less wealth.

The justification for entering the Step 1 and Step 2 variables before the others is straightforward. Step 1 included only variables which describe characteristics one is born with and which are clearly temporally, and thus causally, prior to other variables. Step 2 included only educational variables, and while it is true that some formal education is undertaken in later years, most people have completed this phase by early adulthood. Step 3 contains a diverse set of variables that we do not believe can be arranged in a plausible temporal or causal order. To give one example, household type variables (and marital status) could both cause and be caused by health conditions, by income and by attitudes. Similarly, one's health and attitudes could affect income but the reverse could also be true. If the three-step causal ordering is accepted as correct, or at least plausible, then the estimate that matters for each variable is at the first step in which it enters the analysis. This equation gives the total effect of the variable in question on the dependent variable of household net worth.

Table 11 gives results for prime age households and Table 12 gives parallel results for households of retirement age but not yet retired. Because ageing is so central to wealth accumulation, all equations include 'controls' for age. The equations for working household heads also include an aged squared term to capture the fact that wealth rises with age until about 60 to 65 and then declines as retirement savings are run down.

Table 11: Households headed by persons aged 25 to 54, accounting for differences in net worth, OLS regressions
n=2,867 households Step 1
R2=16.8%
bs
Step 2
R2=18.4%
bs
Step 3
R2=31.5%
bs
Gender: reference person (f=1, m=0) –1.38*** –1.39*** 0.07ns
Age: reference person 0.36*** 0.37*** 0.39***
Age squared (/10) –0.03*** –0.03*** –0.04***
Parents' status (/10) 0.10*** 0.05*** 0.03ns
Australian born 0.41** 0.51*** 0.46***
Other English-speaking born 0.20ns 0.27ns 0.20ns
University degree   0.44** 0.09ns
Trade qualification   0.01ns 0.09ns
Education less than Year 12   –0.56** –0.21***
Couple with no dependants     0.73***
Couple with dependants     0.74***
Lone parent with dependants     –0.33ns
Physical health (/10)     0ns
Mental health (/10)     0.01ns
Work hours (ln)     0.45***
Gross income (ln)     0.38***
Regular saver     0.59***
Irregular saver     0.42***
Takes no risks     –0.82***
Takes average risks     –0.51**
Takes above average risks     –0.42ns

Note: Reference groups: birthplace = non-English speaking; education = Year 12; household type = lone person; saving attitudes = non-saver; risk aversion = takes substantial risks. The equation also included a dummy variable (1–0) for respondents who, in response to the risk aversion question, said they never had any spare cash (so the question of taking risks with money did not arise for them). *** significant at 0.001; ** significant at 0.01; * significant at 0.05; ns not significant.

At Step 1 of the analysis the explanatory variables were the characteristics one is born with. Female-headed households were, as expected, less wealthy than male-headed households. Also in line with expectations, people with high status parents were themselves wealthier than people from lower status backgrounds. Households headed by Australian-born people were on average 46 per cent wealthier than (the reference group of ) households headed by immigrants from non-English speaking countries. Altogether these characteristics of household reference persons accounted for 16.8 per cent of the variance in net worth.

Step 2 of the analysis deals with education, which accounted for another 1.6 per cent of variance. So wealth depends somewhat on education, but rather less so than income. Households headed by people with university degrees were significantly wealthier than the reference group of people who completed Year 12, and people with less than Year 12 education were substantially less wealthy. A separate analysis showed that the education of the household head's partner was also significantly related to total household wealth.19

Step 3 of the analysis—household type—was, as expected, very strongly related to net worth. Couples with and without dependent children (or dependent students) were the two wealthiest types of family. It should be noted that some couples shown in the table as being without dependants had never had children; in other cases their children had grown up and moved out. Lone-parent households were much less well off than couple households, but did not differ significantly from the reference group of one-person households.

Physical and mental health were moderately correlated with wealth (Pearson correlations of 0.08 and 0.14) but neither was statistically significant in this analysis. A more subjective measure of health than the SF-36 scales used here—namely self-reported health on a scale from 'excellent' to 'poor'—was significant, but is arguably a less valid measure of actual health condition.

Next, we consider the effects of the reference person's weekly hours of work and the household's gross income. There was clearly an element of endogeneity involved in including the latter variable, since for some households (for example, shareholders, business owners and property owners), their wealth was a source of income. However, the large majority of non-retired households got their income primarily from paid work. For this group, household income was strongly related to wealth. A separate analysis (not shown) indicated that both the reference person's income and his partner's income were significant separate contributors. The other result shown—that if the reference person worked longer hours, the family was wealthier—is not as obvious as might seem. It actually suggests that, even controlling for income, a family whose head works more generally saves more. May be that is why he/she works long hours.

The savings behaviour and risk aversion variables in Table 11 yield interesting results. Self-reported saving on the part of the household reference person was quite strongly positively related to wealth, while being risk averse was a clear negative.

We now undertake parallel analyses for households headed by people aged 55 to 64. We split this group into those who are still in paid work (Table 12) and those already retired (Table 13).

Table 12: Households headed by persons aged 55 to 64 and not retired, accounting for differences in net worth, OLS regressions
n=461 households Step 1
R2=6.0%
bs
Step 2
R2=8.8%
bs
Step 3
R2=23.2%
bs
Gender: reference person (f=1, m=0) –1.53*** –1.52*** –0.21ns
Age: reference person –0.54ns –0.26ns –1.11ns
Age squared (/10) 0.05ns 0.03ns 0.10ns
Parents' status (/10) 0.10* 0.06ns 0.05ns
Australian born 0.49ns 0.62ns 0.27ns
Other English-speaking born 0.46ns 0.47ns 0.18ns
University degree   0.98* 0.20ns
Trade qualification   0.39ns 0.01ns
Education less than Year 12   –0.21ns –0.23ns
Couple with no dependants     0.26ns
Couple with dependants     0.41ns
Lone parent with dependants     –1.59*
Physical health (/10)     0ns
Mental health (/10)     0.01ns
Work hours (ln)     –0.03ns
Gross income (ln)     0.86***
Regular saver     0.17ns
Irregular saver     0.09
Takes no risks     1.63*
Takes average risks     2.04**
Takes above average risks     2.28**

Note: Reference groups: birthplace = non-English speaking; education = Year 12; household type = lone person; saving attitudes = non-saver; risk aversion = takes substantial risks. The equation also included a dummy variable (1–0) for respondents who, in response to the risk aversion question, said they never had any spare cash (so the question of taking risks with money did not arise for them). *** significant at 0.001; ** significant at 0.01; * significant at 0.05; ns not significant.

Results for non-retired households in this age group in Table 12 should be seen as broadly similar to those for younger working-age households in Table 11. Because of a smaller sample size some relationships which were statistically significant in the previous table are non-significant here, but the effect sizes (regression coefficients) are similar, so it is probably sensible to regard most of the results as much the same.

A few differences are worth drawing attention to. For this cohort, and among those still working, age is no longer positively related to wealth. This is perhaps a little surprising in that one might have expected that, in the case of some reference persons, the very reason for continuing to work would be to increase their wealth prior to retirement. A second difference is more apparent than real. Within this cohort there is no relationship between reported savings behaviour and wealth; but on inspection this turns out to be because almost everyone reported that they were trying hard to save. Finally, there appears to be a non-linear relationship between risk-taking and wealth, with those who take 'substantial risks' (the reference group in this equation) being worse off than the other three groups. However, among the latter groups, those who took 'above average risks' were doing best, while those who took least risks were least wealthy. Again the divergence from the results in Table 11 may be more apparent than real. Only 13 household reference persons in this group reported taking 'substantial risks'. This is too small a sample from which to draw conclusions, but it makes an interesting point. People of this age are the very ones financial advisers suggest should be willing to take more risk in hopes of saving enough to retire. It appears that not many take their advice.20

Now we discuss the retired households in the 55 to 64 years cohort (Table 13).21 Again, most results should be seen as similar to those in Table 11, even if no longer statistically significant due to the small sample size. One difference of some interest is that overseas-born respondents of non-English speaking background were at least as well off as Australians and overseas born of English speaking background. This may indicate that non-Anglo overseas born have made a specially strong effort to save.

Table 13: Households headed by persons aged 55 to 64 and retired, accounting for differences in net worth, OLS regressions
n=318 households Step 1
R2 = 4.0%
bs
Step 2
R2 = 10.9%
bs
Step 3
R2 = 33.1%
bs
Gender: reference person (f=1, m=0) –1.48*** –1.26*** 0.14ns
Age: reference person 0.05ns 0.02ns –0.03ns
Parents' status (/10) 0.09ns –0.03ns –0.05ns
Australian born –0.54ns –0.64ns –0.43ns
Other English-speaking born –0.20ns –0.80ns –0.65ns
University degree   1.99* 1.34ns
Trade qualification   1.48* 1.06ns
Education less than Year 12   –0.15ns –0.07ns
Couple with no dependants     1.64***
Couple with dependants     0.85ns
Lone parent with dependants     1.03ns
Physical health (/10)     0.01ns
Mental health (/10)     0.01ns
Regular saver     0.13ns
Irregular saver     0.04ns
Takes no risks     –0.88ns
Takes average risks     –0.27ns
Takes above average risks     0.30ns

Note: Reference groups: birthplace = non-English speaking; education = Year 12; household type = lone person; saving attitudes = non-saver; risk aversion = takes substantial risks. The equation also included a dummy variable (1–0) for respondents who, in response to the risk aversion question, said they never had any spare cash (so the question of taking risks with money did not arise for them). *** significant at 0.001; ** significant at 0.01; * significant at 0.05; ns not significant.

Table 14 gives a similar analysis for households with heads over 65 years and retired. Because of retirement, working hours are not included and household income is also omitted because in most cases it would directly depend on, rather than contribute to, wealth. The hypotheses to be tested remain the same, except that one would expect wealth to decline with age in retired households (rather than increase) as savings are run down.

Table 14: Retirement age households, accounting for differences in wealth, OLS regressions
n=884 households Step 1
R2=7.7%
bs
Step 2
R2 = 9.8
bs
Step 3
R2 = 23%
bs
Gender: reference person (f=1, m=0) –0.85*** –0.74*** –0.26ns
Age: reference person –0.02** –0.01ns –0.02ns
Parents' status (/10) 0.14*** 0.10** 0.09**
Australian born 0.82*** 0.86** 0.57**
Other English-speaking born 0.22ns 0.21ns –0.03ns
University degree   0.64ns 0.03ns
Trade qualification   0.10ns –0.06ns
Education less than Year 12   –0.39ns –0.41ns
Couple with no dependants     0.42**
Couple with dependants     0.52ns
Physical health (/10)     0ns
Mental health (/10)     0.01ns
Regular saver     0.25ns
Irregular saver     0.06ns
Takes no risks     –0.43ns
Takes average risks     0.13ns
Takes above average risks     0.18ns

Note: Reference groups: birthplace = non-English speaking; education = Year 12; household type = lone person; saving attitudes = non-saver; risk aversion = takes substantial risks. The equation also included a dummy variable (1–0) for respondents who, in response to the risk aversion question, said they never had any spare cash (so the question of taking risks with money did not arise for them). *** significant at 0.001; ** significant at 0.01; * significant at 0.05; ns not significant.

During retirement, as was the case in the working-age population, female-headed households were substantially less wealthy. A second finding is that, as expected, wealth declines with age by a significant although quite small amount (2 per cent a year). This relatively gradual, or slight, decline may suggest that many people choose to keep a fair amount of their wealth intact, perhaps with a view to leaving an inheritance. Parental social status was again fairly strongly related to wealth even in this older age group. And once again the Australian born had greater wealth than non-Anglo immigrants.

Step 3 of the analysis indicated that couple households with no dependants (but not those with dependants) retained a statistically significant wealth advantage over lone-person households in retirement. However, this advantage of about 40 per cent could be regarded as substantively unimportant, given that one-person households require less wealth and income to meet their material needs. Finally, neither the relationship between wealth and savings behaviour, nor that between wealth and risk aversion, were significant in this retired population.

[ Return to Top   Return to Section ]

4. Conclusions and future research need

This report reviews the composition of the assets and debts of Australian households in the last quarter of 2002. It highlights issues relating to the distribution of wealth and to the prospects of mature-age Australians being able to fully or partly self-fund their retirement.

Asset holdings are heavily concentrated in the hands of older households—those within 20 years of retirement and those 10 to 15 years post retirement. This distribution is, to a great extent, due to the fact that asset levels depend on the length of time spent saving and benefiting from the effects of compound interest. It is also a consequence of government policy and initiatives such as the Superannuation Guarantee and generous tax concessions to encourage retirement savings.

Even so, it remains true that the wealth of Australians is still preponderantly in housing. Liquid assets in general, and superannuation holdings in particular, are not yet at an adequate level to enable those currently retired, and most of those approaching retirement, to be entirely self-funding when they finish work. However, many home owner couples are already partly self-funding and more will be in the future. Superannuation holdings are increasing rapidly and are now more widely distributed than in the past. However, the evidence in this report shows that most households within 20 years of retirement are likely to be partly reliant on the Age Pension for their retirement income. These households would have to make significant sacrifices to current living standards to generate enough extra savings to be entirely self-funding.

Current government policy is tackling this problem by changing incentives affecting both the age at which people choose to retire and their likelihood of doing some paid work during retirement. The age at which people can retire with full superannuation entitlements is being gradually raised and a package of measures designed to allow paid work to be combined with receipt of superannuation has been announced. It remains to be seen how effective these changes will be in counteracting the evident desire of most Australians to retire before age 65. A serious underlying problem—confirmed in the most recent surveys—is that most working-age people continue to underestimate the savings they will need to maintain their current lifestyle after they retire (ASFA 2004a).

It should be recognised that all the evidence in this report is cross-sectional—a snapshot. In future research it will be important to gain an understanding of wealth dynamics. It is often assumed that the stock of household wealth, unlike household income flows, is fairly stable and usually just increases gradually over time. Indeed, at a conceptual level, stocks are more-or-less defined as being more stable than flows. Research in Sweden and the United States, however, has shown that, particularly in the former country, asset values have recently been quite volatile (Klevmarken, Lupton & Stafford 2003). All Australian sources agree that, after the early 1990s, the increase in household wealth considerably exceeded the rate of inflation (ABS cat. nos 5232.0 & 5204.0; Kelly 2001; Northwood, Rawnsley & Chen 2003). So it seems quite unlikely that an assumption of wealth stability is justified.

The main assets types are housing, equities and superannuation. Plainly, housing prices, particularly in the capital cities, have increased sharply in recent years and are now starting to fall in real terms. Share prices and managed funds, especially international shares and funds, have also been volatile: having risen rapidly for most of the 1990s, they fell by about 40 per cent in 2000–2002. The value of superannuation assets has also been volatile, since these depend heavily on share prices.

So it is probably mistaken to believe that household wealth is fairly stable. This means it will be important to measure it and assess the causes and consequences of change more frequently than in the past. As the population ages, we need a better understanding of the dynamics of wealth, particularly for those in the retirement and pre-retirement cohorts.

[ Return to Top   Return to Section ]

Appendix A: HILDA household-level wealth derived variables

Appendix A: HILDA household-level wealth derived variables

 

[ Return to Top   Return to Section ]

Appendix B: Assets and debts tables

This appendix provides detailed tables on the assets and debts of households in the cohorts aged 15 to 24, 25 to 34 and 35 to 44. In all cases the households described in this appendix had reference persons who were not retired.

Table B1: The wealth of cohorts, household reference person aged 15 to 24, population weighted estimates
n=347 households Mean
($'000)
Median
($'000)
% of total
assets or debts
% of
households
holding assets/
debts
Overall assets and debts        

Total assets

55.5 12.0    

Total debts

26.9 5.0    

Net worth (assets minus debts)

28.6 5.7    
Assets        

Housing and other property

30.2 0 54.4 17.0(a)

Pensions/superannuation

5.7 2.5 10.3 79.9

Businesses and farms

3.4nr 0 6.1 1.8

Equity investments (shares, managed funds)

2.2 0 4.0 14.7

Bank accounts

3.1 0.7 5.6 92.9

Cars and other vehicles

8.2nr 4.0 14.8 72.3

Other assets(b)

2.7 0 4.9 27.7
  (55.5)   (100.0)  

Non-financial assets

42.2 5.0 76.0 76.0

Financial assets

13.3 5.0 24.0 96.6
  (55.5)   (100.0)  
Debts        

Housing and other property

16.5 0 61.3 15.0

Businesses and farms

1.3nr 0 4.8 0.7

HECS (student) debt

3.6 0 13.4 30.8

Credit (and similar) cards

0.4 0 1.5 25.3

Other debts (cars, hire purchase etc.)

5.0 0 18.6 45.9
  (26.9)   (100.0)  

(a) 14.5 per cent owned the home they lived in (b) Other assets include cash investments, trust funds, the cash-in value of life insurance and collectibles. Note: nr Not reliable—standard error over half the estimate.

Table B2: The wealth of cohorts, household reference person aged 25 to 34, population weighted estimates
n=1,136 households Mean
($'000)
Median
($'000)
% of total
assets or debts
% of households
holding assets/
debts
Overall assets and debts        

Total assets

238.8 140.8    

Total debts

76.2 23.0    

Net worth (assets minus debts)

162.6 74.2    
Assets        

Housing and other property

133.7 9.0 55.9 53.2(a)

Pensions/superannuation

34.5 16.0 14.4 95.9

Businesses and farms

23.4 0 9.8 11.4

Equity investments (shares, managed funds)

8.3 0 3.5 33.8

Bank accounts

9.2 2.4 3.9 96.0

Cars and other vehicles

15.3 10.3 6.4 89.3

Other assets(b)

14.4 0 6.0 37.5
  (238.8)   (100.0)  

Non-financial assets

173.3 6.0 72.6 92.9

Financial assets

65.5 5.7 27.4 99.8
  (238.8)   (100.0)  
Debts        

Housing and other property

63.3 0 83.1 48.2

Businesses and farms

2.5 0 3.3 4.3

HECS (student) debt

1.9 0 2.5 23.5

Credit (and similar) cards

1.2 0 1.6 39.8

Other debts (cars, hire purchase etc.)

7.5 0 9.8 49.0
  (76.2)   (100.0)  

(a) 48.8 per cent owned the home they lived in. (b) Other assets include cash investments, trust funds, the cash-in value of life insurance, and collectibles.

Table B3: The wealth of cohorts, household reference person aged 35 to 44, population weighted estimates
n=1,488 households Mean
($'000)
Median
($'000)
% of total
assets or debts
% of households
holding assets/
debts
Overall assets and debts        

Total assets

450.2 312.3    

Total debts

108.1 60.0    

Net worth (assets minus debts)

342.1 204.9    
Assets        

Housing and other property

256.6 190.0 58.2 75.4(a)

Pensions/superannuation

72.2 30.0 15.7 96.1

Businesses and farms

51.2 0 11.2 18.4

Equity investments (shares, managed funds)

18.9 0 3.9 43.7

Bank accounts

13.5 3.0 3.0 97.9

Cars and other vehicles

20.4 15.0 4.5 94.1

Other assets(b)

17.4 0 3.5 45.2
  (450.2)   (100.0)  

Non-financial assets

324.5 228.0 72.1 96.9

Financial assets

125.7 52.7 27.9 99.9
  (450.2)   (100.0)  
Debts        

Housing and other property

85.1 45.0 78.7 62.5

Businesses and farms

12.3 0 11.4 8.3

HECS (student) debt

0.8 0 0.7 10.6

Credit (and similar) cards

1.4 0 1.3 44.9

Other debts (cars, hire purchase etc.)

8.5 0 7.9 41.7
  (108.1)   (100.0)  

(a) 69.5 per cent owned the home they lived in. (b) Other assets include cash investments, trust funds, the cash-in value of life insurance, and collectibles.

[ Return to Top   Return to Section ]

Endnotes

1. ABS data also have some unavoidable limitations; for example, only financial assets and liabilities are given.

2. The authors also benefited from discussions with ABS and RBA officials who responded generously to questions about inclusions, exclusions and apparent discrepancies between the government sources.

3. This figure was obtained by taking a valuation of $124 billion in Northwood, Rawnsley and Chen (2002, p. 27) for June 2000 and inflating it by 21.4 per cent to obtain an estimate for an average of the September and December quarters of 2002 (21.4 per cent being an estimate of the increase in property values during the period).

4. Overviews of the HILDA data have been previously published in the HILDA annual report for 2003 (Melbourne Institute 2003) and by the Reserve Bank of Australia (RBA 2004). The estimates given in Table 2 differ from both these sources. The HILDA annual report did not include imputed components of wealth. The RBA estimates differ slightly due to omission of imputations for components of wealth included in the HILDA Person Questionnaire, namely bank accounts, superannuation, credit card debt, HECS debt and other personal debt.

5. All results in this paper are weighted to correct for sample bias and attrition.

6. As is usual, the logarithms of both variables were used in order to approximate a normal distribution.

7. Another way to understand the effects of saving is to imagine a country in which all households saved about 3 per cent of income per year and there was no other source of wealth accumulation. This imaginary country would have a wealth distribution similar to that of actual Western countries.

8. For many purposes it is preferable to define the household reference person as the partner with the higher income (plus other criteria). However, in analysing issues relating to retirement and superannuation, this approach can produce anomalies. A not unusual example is a household comprising a recently retired male partner, who previously worked full time and currently receives a modest retirement income, and a female partner receiving a slightly higher part-time working income. This household is best thought of as retired but is classified as non-retired if the higher current earner is treated as the reference person.

9. Multi-family households and group households are not included in Tables 5 and 6.

10. An alternative rule of thumb, which is a little less conservative, says that one needs to save seven times one's household gross income while in paid work in order to achieve a gross income in retirement which is 50 per cent of the pre-retirement income.

11. The household reference person is the male partner in couple-headed households and the single person in one-person households.

12. To be more precise, the limits are different for home owner couples, home owner singles, renter couples and renter singles.

13. The problem with using the mean of the entire cohort is that the estimate is 'distorted' upwards by inclusion of very wealthy households, and so does not give an accurate view of the incomes of typical households. Use of the median has the advantage of providing an accurate view of 'typical' incomes, but the disadvantage that the medians of different components of total income cannot be summed to give the total.

14. 51.3 per cent of this cohort said they had been pressured to retire, compared with 36 per cent of all retired people.

15. No allowance is made for future increases in the real value of the Age Pension during the years when these respondents are retired. If we were to assume, say, a 1.5 per cent increase per year in real values, then the annual incomes shown in columns 5 and 7 would be a little higher.

16. The Pearson correlation was only –0.06 and the standardised regression coefficient was –0.08.

17. If retirement age households had been included, then plainly a large majority of households would be classified as receiving income support, since most receive the Age Pension.

18. SF-36 = 'Medical Outcomes Study 36-Items Short Health Form Health Survey' (see Ware, Snow & Kosinski 2000).

19. If the reference person's partner had a degree, the household was significantly wealthier. However, if the partner had not completed Year 12, this was unrelated to household wealth.

20. The advice is usually given on the basis that children are no longer a major responsibility, so it makes sense to take a chance with higher risk investments.

21. The reason why the 'not significant' figures are low in Tables 13 and 14 is that cases were omitted if missing on any variable in the equations (listwise deletion). The last two attitudinal variables in these tables, and the SF-36 health measures, were in the self-completion questionnaire which has a lower response rate than the main face-to-face questionnaire. It should also be noted that, unlike all results given above, these regression results are unweighted.

[ Return to Top   Return to Section ]

References

Association of Superannuation Funds of Australia (ASFA) 1999, Achieving an adequate retirement income—how much is enough?, ASFA Research Centre, Sydney, October.

—— 2004a, How much do you need to have a comfortable standard of living in retirement?, ASFA, viewed January <www.asfa.asn.au>.

—— 2004b, Westpac–ASFA retirement living standard detailed budget breakdowns: comfortable lifestyle and modest lifestyle, ASFA, viewed February <www.asfa.asn.au>.

Atkinson, AB, Rainwater, L & Smeeding, TM 1995, Income distribution in OECD countries: the evidence from the Luxembourg Income Study, OECD Social Policy Studies no. 18, Organisation for Economic Co-operation and Development, Paris.

Australian Bureau of Statistics (quarterly), Australian National Accounts: Financial Accounts, cat. no. 5332.0, ABS, Canberra.

—— (annual), Australian System of National Accounts, cat. no. 5204.0, ABS, Canberra.

Baekgaard, H 1998, The distribution of household wealth in Australia: 1986 and 1993, NATSEM Discussion Paper no. 34, National Centre for Social and Economic Modeling, University of Canberra, September.

Business Review Weekly 2004, 'Rich 200', May 20–26.

Goode, A & Watson, N (eds) 2007, HILDA User Manual, Release 5.0, Melbourne Institute of Applied Economic and Social Research, University of Melbourne.

Juster, FT, Smith, JP & Stafford, F 1999, 'The measurement and structure of household wealth', Labour Economics, vol. 6, pp. 253–76.

Kelly, S 2001, 'Trends in Australian wealth: new estimates for the 1990s', paper presented at the 30th Annual Conference of Economists, University of Western Australia, Perth, 26 September.

—— 2003, Self-provision in retirement? Forecasting future household wealth, National Centre for Social and Economic Modelling, University of Canberra, December.

Klevmarken, NA, Lupton, JP & Stafford, FP 2003 'Wealth dynamics in the 1980s and 1990s', Journal of Human Resources, vol. 38, pp. 322–53.

Melbourne Institute 2003, HILDA survey annual report 2003, Melbourne Institute of Applied Economic and Social Research, University of Melbourne.

Northwood, K, Rawnsley, T & Chen, L 2002, Experimental estimates of the distribution of household wealth, Australia, 1994–2000, Working Papers in Econometrics and Applied Statistics, Working Paper no. 2002/1.

Piggott, J 1984, 'The distribution of wealth in Australia—a survey', The Economic Record, vol. 60, pp. 252–65.

Podder, N & Kakwani, NC 1976, 'Distribution of wealth in Australia', Review of Income and Wealth, vol. 22, pp. 75–92.

Reserve Bank of Australia 2004, 'The composition and distribution of household estimates and liabilities: evidence from the 2002 HILDA survey', Reserve Bank of Australia Bulletin, pp. 1–11, April.

—— (quarterly), Statement on monetary policy, RBA, Sydney, <www.rba.gov.au/>.

Robertson, G, Grandy, F & McEwin, M 2000, 'Use of social accounting techniques to estimate the distribution of household wealth', paper presented to the Siena Group for Social Statistics Meeting, Maastricht, The Netherlands, May 23–26.

Ware, JE, Snow, KK & Kosinski, M 2000, SF-36 Health Survey: Manual and interpretation guide, QualityMetric Inc., Lincoln, RI.

Watson, N & Wooden, M 2002, The Household Income and Labour Dynamics in Australia (HILDA) survey: Wave 1 survey methodology, HILDA Project Technical Series no. 1/02, Melbourne Institute of Applied Economic and Social Research, University of Melbourne.

Webster, E 2000, 'The growth of enterprise intangible investment in Australia', Information Economics and Policy, vol. 12, pp. 1–25.

Yates, J 1991, Australia's owner-occupied housing wealth and its impact on income distribution, Reports and Proceedings no. 92, Social Policy Research Centre, University of New South Wales, Sydney.

[ Return to Top   Return to Section ]

List of tables

Table 1: Aggregate wealth sources, summary of differences in scope

Table 2: Overview: assets, debts and net worth of households in 2002, population weighted estimates

Table 3: The distribution of wealth and income, Gini coefficients

Table 4: Shares of total wealth (net worth) by deciles

Table 5: Age cohorts, net worth of households at the cohort mean and at the 10th, 30th, median, 70th and 90th percentiles

Table 6: Financial assets and incomes of retired couple home owners, cohorts aged 60-64, 65-69, 70-74 and 75+, population weighted results

Table 7: Projected financial assets and income, retirement at 60 or 65, population weighted results

Table 8: Financial assets and incomes of couple renter households and households headed by non-partnered people, population weighted results

Table 9: Assets and debts of income support recipients, household reference person receiving income support and not retired

Table 10: The assets and debts of lone-parent households

Table 11: Households headed by persons aged 25 to 54, accounting for differences in net worth, OLS regressions

Table 12: Households headed by persons aged 55 to 64 and not retired, accounting for differences in net worth, OLS regressions

Table 13: Households headed by persons aged 55 to 64 and retired, accounting for differences in net worth, OLS regressions

Table 14: Retirement age households, accounting for differences in wealth, OLS regressions


Appendix tables

Table B1: The wealth of cohorts, household reference person aged 15 to 24, population weighted estimates

Table B2: The wealth of cohorts, household reference person aged 25 to 34, population weighted estimates

Table B3: The wealth of cohorts, household reference person aged 35 to 44, population weighted estimates

[ Return to Top   Return to Section ]

Content Updated: 20 August 2013