Data Surveys Measures of Household Wealth for Australia

Transcription

1 The Australian Economic Review, vol. 42, no. 2, pp Data Surveys Measures of Household Wealth for Australia Paul Bloxham and Thomas Betts Economic Analysis Department, Reserve Bank of Australia 1. Introduction Over the past two decades, the nominal value of household wealth in Australia has more than quadrupled. This has been part of an expansion of the household sector s balance sheet, which has been a structural feature associated with financial deregulation and rising incomes and has occurred in a number of developed economies. Given these trends, and an expanding theoretical literature on the relationship between wealth accumulation and savings behaviour, there has been increased interest in robust measures of household wealth. This has prompted a significant increase in the quantity and quality of household wealth data for Australia over recent years. This article describes and compares the four main measures of household wealth for Australia. These are: A quarterly time series measure constructed by the Reserve Bank of Australia (henceforth, the RBA measure); An annual time series measure produced by the Australian Bureau of Statistics (ABS) as part of the compilation of the Australian System of National Accounts (henceforth, the ASNA measure). This measure and the RBA measure use source data that overlap; This article uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA), Australian Government, and is managed by the Melbourne Institute of Applied Economic and Social Research. The findings and views reported in this article, however, are those of the authors and should not be attributed to FaHCSIA, the Melbourne Institute, or the Reserve Bank of Australia. A measure of the distribution of wealth based on the Survey of Income and Housing by the ABS, which is an independent crosssectional survey of households containing a range of questions regarding households balance sheets (henceforth, the SIH measure); and A measure of the distribution of wealth based on the Household, Income and Labour Dynamics in Australia Survey, which is a longitudinal survey of households that periodically contains a module of questions on household wealth (henceforth, the HILDA measure). 1 Section 2 of this article describes the various components of household wealth and compares the different wealth measures in aggregate. Section 3 describes and compares the sources of the data used to construct the tangible assets component of household wealth and Section 4 does the same for the financial assets and liabilities components. Section 5 discusses the available information on the distribution of wealth across households, some recent trends in these data and some of the issues associated with the use of the panel aspect of the HILDA Survey. Section 6 concludes. 2. Aggregate Household Wealth For all of the measures described here, household wealth is the sum of households tangible and financial assets minus households liabilities (human capital and other intangible assets, such as patents, are excluded). Households tangible assets comprise real estate assets and consumer durables and account for around three-fifths of the value of the total household assets in Australia, which is high relative to Published by Blackwell Publishing Asia Pty Ltd

2 218 The Australian Economic Review June 29 Table 1 Household Wealth Level (A$ billion) Share of assets (%) Assets Tangible Real estate Durable goods Financial a Superannuation and life offices Currency and deposits Equities Other Total Liabilities Total b Wealth (assets less liabilities) Financial a,b Total a,b Memo: Total as a share of GDP (%) Notes: (a) Includes the financial assets of unincorporated enterprises. (b) Includes the liabilities of unincorporated enterprises. Sources: ABS;RBA. other comparable countries (for a discussion, see Ellis and Andrews 21). 2 Households financial assets comprise mostly holdings with superannuation funds and life offices, deposits with financial institutions and direct holdings of listed equities. The rise in the value of households assets over recent years has been accompanied by a significant increase in household liabilities, which predominantly comprise loans used to purchase housing. Table 1 shows the estimates of the components of household wealth, based on the RBA measure, to provide a guide as to the absolute and relative sizes of the components of household wealth in Australia. All four available measures of household wealth provide broadly similar estimates of the value of aggregate household assets, liabilities and wealth in comparable periods, with a large part of any remaining variation across the measures explained by differences in the scope of the surveys, rather than by measurement error (Figure 1). 3 The differences in scope mainly arise due to the coverage of different types of assets and liabilities. These, along with differing levels of disaggregation in the available surveys, limit the extent of direct comparisons between equivalent measures. For example, the treatment of unincorporated enterprises and not-for-profit institutions serving households (NPISHs), which are often hard to distinguish from households, differs across the measures. The ASNA measure includes unincorporated enterprises and NPISHs. The RBA measure includes the financial assets of unincorporated enterprises and NPISHs but excludes their tangible assets and financial liabilities. In the HILDA Survey and SIH, households holdings of unincorporated business assets and liabilities are collected and separately identified, so they can be (and for our purposes are) excluded in the household wealth measures. The differences in survey scope and coverage of the components of household wealth are discussed in more detail below. The RBA and ASNA measures of household wealth are largely based on data from financial institutions, collected via surveys by the ABS and other authorities. For the RBA time series wealth measure, quarterly estimates are available from June quarter 1988, which is when the ABS quarterly Financial Accounts data begin, while annual estimates from the ASNA are available from A number of estimates of household wealth for the period

3 Bloxham and Betts: Measures of Household Wealth for Australia 219 Figure 1 Household Assets and Debt 8 RBA SIH Assets 6 ASNA HILDA Debt Wealth Year Notes:TheAustralian System of National Accounts (ASNA) includes unincorporated enterprises and not-for-profit institutions serving households; HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. prior to 1989 are available (see Callen 1991 and Goldbloom and Craston 28), although with no data available on the household sector holdings of business sector assets for this period, these measures rely on grossed-up estimates of the value of business sector assets and attribution of ownership to the household sector. Here, we focus on the data since June quarter 1988, as it is from this period that comprehensive data on the financial assets of households are available, via the ABS Financial Accounts. The HILDA and SIH measures of wealth are based on periodic, direct surveys of households. The HILDA Survey s wealth module commenced in 22, covers around 75 households and, although the core survey is conducted annually, detailed information on households financial positions is only collected in the quadrennial wealth module. The SIH is a survey of around 1 households. Since 23 4, it has contained detailed information on households balance sheets, with the most recent data being available for The next collection is scheduled for 29 1 and then every six years afterwards. Prior to this, the ABS had conducted a number of surveys of household finances since 1982,

4 22 The Australian Economic Review June 29 including the Household Expenditure Survey, but the only wealth components covered were housing assets and mortgage debt. Estimates of the distribution of total household wealth for periods prior to 22 are only available by imputation from the surveyed data on income streams from assets (see Bækgaard 1998 and Kelly 21 for examples of this approach). 3. Tangible Assets The bulk of the value of tangible assets comprises real estate holdings. The measures of real estate assets include both the value of the land and the structures. Durable goods account for the remaining tangible assets and, as can be seen in Table 1, are only a small share of household sector wealth. 3.1 Real Estate Assets The RBA measure of real estate assets includes residential property only and is estimated as the product of the number of dwellings owned by households and the mean dwelling sales price. The number of dwellings owned by households is calculated using data from the five-yearly Census, with quarterly estimates interpolated between the census years using the volume of private residential building completions (ABS 28b) and an estimate of demolitions. The dwellings owned by the government and corporate sector are excluded (the Census data suggest these sectors owned 6.5 per cent of all dwellings in 26). As the number of dwellings grows at a much more stable rate than the average price, time series variation in the estimate of residential property assets is largely driven by changes in prices (Figure 2). The dwelling price measure is the seasonally adjusted mean sale price of the residential properties transferred each quarter. This price is derived from a sample of property transactions sourced from a private sector provider, Australian Property Monitors (APM), since 1993 (which are cleaned and seasonally adjusted by the RBA) and uses data from the Commonwealth Bank (CBA) prior to this time. Australian Property Monitors constructs its series using data from the State Valuer General (VG) offices, which are contract-dated and eventually cover all transactions. 4 The price data cover all of the housing in the capital cities and regional areas, except Tasmania and the regional Northern Territory, and rural and commercial land are not included. The house price data from the CBA are from its loan book and have a similar geographical coverage to the APM measure. To aggregate the dwelling price measure, weights are applied at the capital city and rest-of-state level based on the proportion of all dwellings in those locations, as measured by the Census. The ASNA measure of household real estate assets uses the RBA measure of dwelling assets and adds estimates of the value of commercial and rural land owned by the household sector, in line with the ASNA measure s coverage of unincorporated enterprises and NPISHs (ABS 26a). The ABS obtains estimates of the total value of commercial and rural land from the State VGs and apportions them to the various institutional sectors: the household sector is apportioned 8 per cent of rural land and 2 per cent of commercial land (ABS 26b). In contrast to the RBA and ASNA measures, both the HILDA and the SIH measures of dwelling wealth are based on the subjective judgement of households about the value of their property assets, as reported in face-to-face interviews. Despite the different approaches, the four measures provide broadly similar estimates of the aggregate value of real estate assets (Figure 3). As would be expected, the ASNA measure is higher than the RBA measure due to its broader scope. The HILDA-based and SIHbased measures also suggest a higher level of household real estate assets than the RBA measure in comparable periods (both are around 7 per cent higher), which might reflect overestimates by households of their own home value, given the general rise in dwelling prices in the lead up to those periods. 5 To some extent, it also could reflect bias in the measure of house prices used in the construction of the RBA and ASNA wealth measures, as the houses that are sold in any period might not be representative of the overall housing stock. However, this effect could work in either direction and the

5 Bloxham and Betts: Measures of Household Wealth for Australia 221 Figure 2 Value of Dwellings (year-ended percentage change) 4 3 Dwelling prices Number of dwellings 2 % Year Sources: ABS; Australian Property Monitors; RBA. Figure 3 Household Real Estate Assets RBA ASNA SIH HILDA Year Notes:TheAustralian System of National Accounts (ASNA) includes unincorporated enterprises and not-for-profit institutions serving households; HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. authors are unaware of any empirical work for Australia that establishes the likely direction of any bias Household Durable Goods Durable goods accounted for around 5 to 1 per cent of household assets in 26, depending on what is included in the definition of durables and the valuation technique applied. Figure 4 shows the data from the four different measures, as at 26. The value of durables in the RBA, ANSA and HILDA measures are similar, while the value in the SIH measure is significantly higher. The RBA and ASNA measures of consumer durables both include motor vehicles and household furnishings and equipment and are constructed using the data from the ABS national accounts, which uses the perpetual inventory model to calculate the stocks

6 222 The Australian Economic Review June 29 Figure 4 Household Durable Goods (26) Notes: ASNA denotes the Australian System of National Accounts; HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. from the value of flows of new durables (adjusted for an assumed depreciation rate). The ANSA measure does not allow separate identification of the types of durables. The HILDA Survey includes collectibles, such as artworks, coins, jewellery, stamps, vintage cars and wine, but excludes other home contents and is based on the reported re-sale value of these items. The SIH measure differs from the others due to the valuation technique applied, as well as its scope. The SIH includes the same items as the RBA and ASNA measures, as well as collectibles, but is based on the households reported insured value of these durables, where available. 4. Financial Assets and Liabilities The bulk of the financial assets of Australian households is held with Australian financial institutions in the form of superannuation and deposits. Direct equity holdings account for less than 1 per cent of household assets, though households also have exposure to equity markets indirectly through their holdings of superannuation assets. The bulk of households debt is owed to domestic financial institutions and is used to purchase housing. 4.1 Financial Assets The RBA and ASNA measures of household financial assets are both based on data compiled by the ABS and published in the quarterly Financial Accounts (ABS 28c). The Financial Accounts estimates are largely sourced from data collected from financial institutions using two quarterly ABS surveys: the Survey of Financial Information and the Survey of International Investment. This information is combined with the data obtained from the RBA, the Australian Prudential Regulation Authority (APRA) and other sources. In contrast, the HILDA Survey and SIH contain questions for households about their financial position. During the surveys conducted as part of the SIH, the interviewer is instructed to request that the surveyed households refer to financial statements, such as bank balances and equity certificates, to verify the responses. The estimates of the value of financial assets vary across the surveys. Figure 5 shows the total financial assets measures, as well as the main components. Some of the differences in the estimates can be explained by scope and coverage (Table 2 provides some details). One key difference for the estimates of superannuation

7 Bloxham and Betts: Measures of Household Wealth for Australia 223 Figure 5 Household Financial Assets Financial accounts Financial accounts (excluding unfunded super) SIH HILDA Total 1.6 Superannuation Currency and Deposits Equities Other Financial Year Notes: Equities includes the equity held in unlisted unincorporated enterprises and not-for-profit institutions serving households; HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing.

8 224 The Australian Economic Review June 29 Table 2 Treatment of Selected Financial Assets in the Household Wealth Measures Asset RBA ASNA a SIH b HILDA c Unfunded superannuation Included and separately identified Largely excluded Largely excluded Self-managed superannuation funds Included as superannuation Included as equities Included as superannuation Unlisted companies Excluded Separately identified Included equity (and excluded here) Private trusts Classified by asset type Included in other financial assets Included in other financial assets Cash management trusts Included in currency and deposits Included in trusts Included in trusts Notes: (a) ASNA denotes the Australian System of National Accounts. (b) SIH denotes the Survey of Income and Housing. (c) HILDA denotes the Household, Income and Labour Dynamics in Australia Survey. assets is the treatment of unfunded superannuation assets largely defined benefit superannuation assets of public sector employees. The unfunded superannuation assets are included (and separately identified) in the ABS Financial Accounts, but are largely excluded from the HILDA Survey and SIH. Even after adjusting for this, the estimates of superannuation assets in the SIH are significantly lower than those in the other wealth measures. This could reflect the treatment of the equity holdings of self-managed superannuation funds in the SIH as equities, rather than superannuation assets, and also would help to explain why the SIH has higher estimates of equity holdings than the other surveys. Another reason for the different estimates of equity holdings across the surveys is the different treatment of shares in listed and unlisted companies and of private trusts (included in other financial assets in Figure 5). In particular, the HILDA measure includes some unincorporated equity holdings, while the SIH measure does not. The ABS Financial Accounts also allocates the assets held in private trusts according to the type of underlying asset, while the household surveys account for these assets as a separate trust asset class. The estimates of households holdings of currency and deposits vary significantly between the aggregate Financial Accounts data and the HILDA and SIH measures: over recent years, the Financial Accounts estimates generally have been A$15 billion to A$2 billion higher than the survey-based estimates. This is partly due to the inclusion of cash management trust assets in the Financial Accounts classification of currency and deposits, with household survey respondents likely to count these as trust assets. Nonetheless, this alone cannot account for the substantial difference and it is likely to at least partly reflect underreporting by the surveyed households (see ABS 27). In the United States, large differences between the estimates of currency and deposits from the Federal Reserve s Flow of Funds Accounts and the Survey of Consumer Finances are partly attributed to under-reporting by the surveyed households (see Antoniewicz 2). 4.2 Liabilities The RBA measure of household debt is largely based on balance sheet data collected by the APRA from Authorised Deposit Taking Institutions in Australia. The household sector estimates are derived from on-balance sheet information on housing loans and other personal loans (including margin loans, credit cards and other fixed loans) to households, as well as data on securitised loans. The RBA estimates exclude loans to unincorporated enterprises and NPISHs. The ASNA measure is also derived from financial institution data but includes the liabilities of unincorporated enterprises and NPISHs, loans to households from the government sector, such as the Higher Education Contribution Scheme (HECS) debt, and

9 Bloxham and Betts: Measures of Household Wealth for Australia 225 Figure 6 Household Debt (26) Notes: ASNA denotes the Australian System of National Accounts; HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. other financial obligations of households, such as tax payable and bills of exchange. The ASNA measure does not allow separate identification of the different types of debt. For both the HILDA and SIH measures, total household debt is the sum of property loans, HECS debt, outstanding balances on credit cards, loans for vehicle purchases and investment loans or loans for other purposes; they exclude the debts of unincorporated businesses owned by households. The ASNA measure of household liabilities is higher than the other three estimates, largely due to the inclusion of the liabilities of unincorporated enterprises and NPISHs (Figure 6). The RBA measure, which excludes these entities, is still higher than the two survey-based measures. The largest difference between the measures is due to the estimates of housing debt. The lower levels of housing debt in the household surveys might reflect a general tendency of households to understate the value of their debt when responding to household surveys. 7 To a small extent, the lower estimates in the household surveys also could reflect an underestimate of credit card debt, as households that just use credit cards as a means of payment might not report this as debt (Headey 23). 5. The Distribution of Household Wealth The HILDA Survey and SIH provide detailed information about the distribution of household wealth across a range of different dimensions (the SIH contains around 4 data items while the HILDA Survey contains over 4 data items). Geographically, the HILDA Survey provides data down to the statistical local area level, of which there are around 14 in Australia, while the SIH can separately identify households in capital cities and the rest of the state. Here, we compare the SIH and HILDA distribution of wealth measures across some commonly assessed dimensions wealth, income and state. 8 We then discuss the longitudinal aspect of the HILDA Survey. 5.1 The Distribution of Household Wealth across a Number of Dimensions One interesting dimension across which to examine the distribution of wealth is by wealth quintiles. Data from both the SIH and HILDA Survey show similar results across this dimension (Figure 7). According to both surveys, the mean value of wealth for the top quintile (the wealthiest 2 per cent of households) is around threefold that of the second-highest quintile and

10 226 The Australian Economic Review June 29 Figure 7 Distribution of Household Wealth (mean) 2 15 SIH (25 6) HILDA (26) $' 1 5 1st 2nd 3rd 4th 5th Wealth quintile Notes: HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. around 1-fold that of the second-lowest quintile. In dollar terms, average household wealth is under A$3 per household for the lowest quintile and is around A$1.7 million for the highest quintile. The implied Gini coefficient for the distribution of wealth across households is.59 in the 25 6 SIH and.61 in the 26 HILDA Survey. When we rank households by income quintile, we also find a positive skew in the wealth distribution of households, with the skew apparent across the distribution of all of the major components of wealth (Figure 8). Both the HILDA Survey and SIH show this, though there are some small differences between the estimates at different points in the income distribution. In 26, the mean value of non-financial assets in the highest income quintile was threefold-to-fourfold the mean of the lowest income quintile, for both the SIH and HILDA measures, while the mean value of financial assets in the highest income quintile was 7 1-fold that of the value of the lowest income quintile. Likewise, the households with the highest levels of debt were those at the upper end of the income distribution: household debt in the top income quintile was 1-fold that of the bottom income quintile. Households in the lowest income quintile had close to no debt, which is unsurprising as these households are unlikely to have much access to finance. The data from the SIH and HILDA Survey suggest that the majority of the variation in household wealth across the mainland states can be explained by differences in the value of households real estate assets (Figure 9). In contrast, the estimated levels of financial assets are relatively similar across the states. The relative debt levels across the states are positively correlated with the average dwelling asset levels, which is consistent with the fact that a large part of household debt is used to finance residential property purchases. 5.2 Panel Aspects of the Data Using the various cross-sections from different periods, it is possible to examine how the distribution of wealth has changed over time. Though this can be done using either the SIH or HILDA data, a key advantage of the HILDA data is that they are longitudinal (panel) data, which allows survey respondents to be matched through time. Of the households who responded to the wealth module in the 22 survey, around four-fifths of the same households also responded to the 26 survey (note that the HILDA Survey tracks individuals,

11 Bloxham and Betts: Measures of Household Wealth for Australia 227 Figure 8 Household Balance Sheets (mean, by gross income quintile) 1 Non-Financial Assets Financial Assets Debt $' SIH (25 6) HILDA (26) 2 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th Gross income quintile Notes: HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. Figure 9 Household Balance Sheets (mean, by state) 1 Non-Financial Assets Financial Assets Debt 8 $' 6 4 SIH (25 6) HILDA (26) 2 NSW Vic Qld SA WA NSW Vic Qld SA WA NSW Vic Qld SA WA Notes: HILDA denotes the Household, Income and Labour Dynamics in Australia Survey; SIH denotes the Survey of Income and Housing. although the questions refer to the household as a unit). The longitudinal aspect of the HILDA Survey allows researchers to tackle questions that other data sources cannot address. In fact, simply looking at the changes in key aggregates from independent cross-sections often can mask more complex dynamics in the distribution and composition of household wealth. For example, while the aggregate data suggest that household wealth increased strongly between 22 and 26 and the independent crosssectional data suggest median wealth grew strongly in each of the quintiles over this period, the panel data reveal that wealth fell for some households. To put some numbers on it, the HILDA Survey suggests that mean household wealth increased by 5 per cent between 22 and 26 and all quintiles had growth in their mean wealth of at least 4 per cent. However,

12 228 The Australian Economic Review June 29 one-fifth of households appear to have reported a fall in their household wealth, which reflects households moving from one part of the wealth distribution to another, albeit measured with some error (discussed below). This result is only revealed using the panel data, as they allow the analyst to stratify the data by changes between periods (for example, growth in household wealth), while independent crosssections only allow the analyst to group data by levels (for example, the level of household wealth). In this way, the panel data allow the analyst to see more complex changes and to develop behavioural explanations for these trends by examining the characteristics of individual households matched through time, such as the labour market status or the age of the household head, between the periods. For example, falls in wealth might be related to retired households running down their wealth, consistent with the life cycle hypothesis, or household members becoming unemployed and drawing down their wealth in place of diminished income. When using the panel data, however, there are a number of measurement issues that need to be considered. Here, we highlight three of them. First, there might be sample attrition bias: households that drop out of the sample might not be representative of the population. For example, households with unfavourable wealth outcomes might be less inclined to be repeat respondents. This could bias the results obtained from the remaining sample of households. Second, rounding might cause errors, particularly for households with lower wealth. For example, households might round their responses, to say, the nearest $1. This would cause a larger relative measurement error for lower-wealth households than for those with higher wealth. Third, the growth estimates across the distribution might be biased by misreporting. For example, a low-wealth household that misreports a high level of wealth in the first period a common case would be adding an extra zero to the reported wealth but reports correctly in the second period, will be (incorrectly) in the bottom tail of the distribution of wealth growth. This misreporting is likely to have its largest effects on the tails of the distribution of wealth growth (for a discussion, see Wilkins, Warren and Hahn 29). In all, although the longitudinal data provide a rich source of information, measurement issues require that caution is applied when interpreting these data. 6. Conclusion There has been a significant increase in the quantity and quality of the measures of household wealth available for Australia in recent years. By international standards, Australia now has a rich set of household wealth data, particularly as it now has a longitudinal survey of household wealth, which is quite rare. In this article, we describe and compare the four main measures available, which include two time series measures the RBA and ANSA measures and two measures of the distribution of wealth the HILDA and SIH measures. The different measures of wealth are clearly useful for different purposes and, as would be expected, the time series measures have a narrower coverage than the distributional measures, but have the advantage of better comparability over a relatively long time history. The distributional measures allow more detailed comparisons to be made across other dimensions, including by wealth, income and geographical location. We have shown that, in aggregate, the four measures are broadly similar and that the differences that do occur tend to reflect the differences in the scope of the measures. This should provide researchers with some confidence about the reliability of the data. February 29 Endnotes 1. The Reserve Bank of Australia (RBA) measure is published in the monthly Bulletin by the RBA, in Table B2. The ASNA measure is published annually in the Australian System of National Accounts (ABS 28a). Access to the Household, Income and Labour Dynamics in Australia Survey datasets can be obtained from the Melbourne Institute (< The results from recent surveys of income and housing are available in the Household Wealth and Wealth Distribution publication (ABS 27) and access to the unit record data can be obtained from the ABS.

13 Bloxham and Betts: Measures of Household Wealth for Australia For the time series data on aggregate household assets and liabilities for a range of developed countries, see the Organisation for Economic Co-operation and Development s (28) annex tables, which are updated biannually. 3. At the national level, the four aggregate measures of wealth differ by no more than 8 per cent. By comparison, the differences across the surveys for the United States are larger and are generally thought to arise because of problems measuring the assets of the very wealthy (top 1 per cent of the distribution), which has prompted oversampling of the upper end of the wealth distribution in the US Survey of Consumer Finances (SCF) (Juster, Smith and Stafford 1999). The differences between the wealth estimates from the SCF and another US household survey, the Panel Study of Income Dynamics, have been up to 25 per cent. 4. For the latest couple of quarters, the data are incomplete, as it takes time to process the transactions. To improve the representativeness of the sample, Australian Property Monitors supplements the Valuer General data with data collected from real estate agents and other sources. In general, more expensive dwellings take longer to settle, so estimates for the most recent quarters are biased downwards. To correct this bias, the Reserve Bank of Australia dwelling assets measure is constructed using a stratified median measure of dwelling prices for the two latest quarters only. For details on this measure, see Prasad and Richards (26). 5. Benítez-Silva et al. (28) show that, in the US surveys, individuals tend to overstate the value of their own home and that this varies with broader economic conditions. 6. For example, if rental properties are lower-quality than owner-occupied properties and are turned over more often than owner-occupied properties, this could downward bias the estimates of house sale prices. However, it is also feasible that higher-quality properties are turned over more often than lower-quality properties, particularly as the sale price data include the sales of new homes. 7. A similar result is found in the United States, where the aggregate debt estimates from the Federal Reserve s Flow of Funds Accounts are larger than those obtained from the Survey of Consumer Finances Household Survey for comparable periods. 8. For a more comprehensive examination of the Household, Income and Labour Dynamics in Australia Survey data, see Kohler, Connolly and Smith (24), Headey, Marks and Wooden (24) and Bloxham and Betts (29). References Antoniewicz, R. 2, A comparison of the household sector from the Flow of Funds Accounts and the Survey of Consumer Finances, Federal Reserve Board Working Paper, Washington, DC, viewed March 29,< workingpapers.htm>. Australian Bureau of Statistics 26a, Australian System of National Accounts, 25/6, Feature Article: Household Sector Balance Sheet A National Accounts Perspective, Cat. no. 524., ABS, Canberra, viewed March 29,< gov.au>. Australian Bureau of Statistics 26b, Australian System of National Accounts, 25/6, Feature Article: Valuing Land and Dwellings Owned by Households, Cat. no. 524., ABS, Canberra, viewed March 29, < Australian Bureau of Statistics 27, Household Wealth and Wealth Distribution, Cat. no , ABS, Canberra, viewed March 29, < Australian Bureau of Statistics 28a, Australian System of National Accounts, 27/8, Cat. no. 524., ABS, Canberra, viewed March 29, < au>. Australian Bureau of Statistics 28b, Building Activity, Australia, Cat. no , ABS, Canberra, viewed March 29, < Australian Bureau of Statistics 28c, Financial Accounts, Cat. no , ABS, Canberra, viewed March 29, < Bækgaard, H. 1998, The distribution of household wealth in Australia: 1986 and 1993, National Centre for Social and Economic Modelling Discussion Paper no. 34, Canberra. Benítez-Silva, H., Eren, S., Heiland, F., Jiménez-Martin, S. 28, How well do individuals predict the selling prices of their homes?, Fundación de Estudios de Economía Aplicada Working Paper 28-1, Madrid. Bloxham, P. and Betts, T. 29, The composition and distribution of household wealth in Australia, Reserve Bank of Australia Bulletin, April, viewed April 29, < Callen, T. 1991, Estimates of private sector wealth, Reserve Bank of Australia Research

14 23 The Australian Economic Review June 29 Discussion Paper 919, Sydney, viewed March 29, < Ellis, L. and Andrews, D. 21, City sizes, housing costs, and wealth, Reserve Bank of Australia Research Discussion Paper 21-8, Sydney, viewed March 29, < Goldbloom, A. and Craston, A. 28, Australian household net worth, Treasury Economic Roundup, Summer, pp. 51 8, viewed March 29, < gov.au>. Headey, B. 23, Income and wealth Facilitating multiple approaches to measurement and permitting different levels of aggregation, HILDA Project Discussion Paper Series no. 3/3, Department of Family and Community Services, Australian Government. Headey, B., Marks, G. and Wooden, M. 24, The structure and distribution of household wealth in Australia, Melbourne Institute of Applied Economic and Social Research Working Paper no. 12/4, University of Melbourne. Juster, F. T., Smith, J. P. and Stafford, F. 1999, The measurement and structure of household wealth, Labour Economics, vol. 6, pp Kelly, S. 21, Trends in Australian wealth New estimates for the 199s, paper presented to 3th Annual Conference of Economists, Perth, September. Kohler, M., Connolly, E. and Smith, K. 24, The composition and distribution of household assets and liabilities: Evidence from the 22 HILDA Survey, Reserve Bank of Australia Bulletin, April, pp. 1 11, viewed March 29, < gov.au>. Organisation for Economic Co-operation and Development 28, Annex tables, OECD Economic Outlook, no. 84, viewed March 29, < Prasad, N. and Richards, A. 26, Measuring housing price growth Using stratification to improve median-based measures, Reserve Bank of Australia Research Discussion Paper 26-4, Sydney, viewed March 29, < Wilkins, R., Warren, D. and Hahn, M. 29, Families, Incomes and Jobs, Volume 4: A Statistical Report on Waves 1 to 6 of the HILDA Survey, Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Melbourne, forthcoming. Copyright and Disclaimer Notices HILDA Survey TheHousehold, Income and Labour Dynamics in Australia (HILDA) Survey was initiated and is funded by the Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA), Australian Government, and is managed by the Melbourne Institute of Applied Economic and Social Research. The findings and views based on these data should not be attributed to either FaHCSIA or the Melbourne Institute. Australian Property Monitors Copyright 24 Australian Property Monitors Pty Ltd. The particular state and territory governments hold copyright in the government-sourced data. The data is used with permission under licence. No liability is accepted. In compiling this information, Australian Property Monitors (APM) relies upon information supplied by a number of external sources. The information is supplied on the basis that, although APM believes that all the information will be correct at the time of publication, it does not warrant its accuracy or completeness. Queensland: The underlying data for Queensland are copyrighted to the State of Queensland, for which no responsibility is accepted. South Australia: The copyright in this information belongs to the South Australian Government, which does not accept any responsibility for the accuracy or completeness of the information or its suitability for any purpose. Victoria: Some underlying Victorian information is copyright of the State of Victoria (Department of Sustainability and Environment). The accuracy is not warranted. Its use is on the basis that the State of Victoria shall bear no responsibility or liability whatsoever for any errors, faults, defects or omissions in that information.

15 Bloxham and Betts: Measures of Household Wealth for Australia 231 Western Australia: Copyright of the State of Western Australia (Department of Land Administration, or DOLA) (24). Licence no. PA Based on information provided with the permission of DOLA. Australian Capital Territory: These data are the property of the Australian Capital Territory. No part of them may, in any form or by any means (electronic, mechanical, microcopying, photocopying, recording or otherwise), be reproduced, stored in a retrieval system or transmitted without prior written permission. Enquiries should be directed to: The Executive Director, ACT Planning and Land Management, GPO Box 198, Canberra, ACT 261.