RDP 2001-08: City Sizes, Housing Costs, and Wealth 2. Dwellings and Household Wealth

2.1 Measuring Dwelling Wealth

Measurement of the value of the stock of dwelling wealth is in principle as simple as counting the number of dwellings, and multiplying by an appropriately weighted estimate of the average prices of those dwellings. The first part can be generated in a straightforward way using national census data, and interpolated between census dates using information on dwelling completions. The second part, which must be available in local currency values rather than as an index number (as is the case with the ABS House Price Index), is more difficult to obtain.

Some statistical agencies publish estimates of the value of the dwelling stock as part of the country's national accounts. However, national accounting principles do not capture the market price of dwellings, including land value, which is what matters for household wealth. Similarly, implicit price deflators for dwelling investment from national accounts do not correspond to market prices of the existing dwelling stock because they generally exclude land and are based on the composition of new dwellings, not the stock of existing dwellings. Price deflators also exclude the effects of increasing dwelling quality, for example where new houses are larger on average than those built previously. However, these effects clearly add to households' dwelling wealth.

The most appropriate sources for data on the market value of dwellings are those based on prices of existing dwellings sold. These series are sometimes collected by national statistical agencies but are more likely to be published by financial institutions or real estate associations. Dwelling prices are frequently used as an indicator of more general price pressures (Girouard and Blöndal 2001). Therefore many published series on sale prices abstract from compositional effects, or relate only to specific markets or types of housing – for example, detached houses in major urban areas, houses for which past sale prices are known, or dwellings of a standardised size. These adjustments ensure that the series are close to a pure price signal, but are unhelpful when trying to determine the market value of the total dwelling stock.

Appropriate measures of the market price of dwellings must include all regions of the country, and apartments and townhouses as well as detached houses. For this reason, the Reserve Bank uses data from the Housing Industry Association's Housing Report, based on prices paid by customers of the Commonwealth Bank. Unlike the other dwelling price series available in Australia such as those from the Real Estate Institute of Australia (REIA) and the ABS, these data include all dwellings, not just detached houses, and cover non-metropolitan regions. However, we are therefore implicitly assuming that Commonwealth Bank customers are representative of all home-buyers. The CBA/HIA prices tend to be higher than those reported by the REIA; it is difficult to say which is correct, given that the CBA/HIA series is otherwise conceptually superior. However, if the CBA/HIA data did overstate the true level of housing prices in Australia, we would have a smaller distance from the dwelling wealth levels of other countries to explain.

Even when conceptually correct measures of sale prices are available, some aspects of their construction can still create biases in estimates of dwelling stock values. Measured values can be biased down by the use of median rather than average prices, for what is likely to be a left-skewed distribution. There is also a potential bias in average dwelling-price measures if different types of dwellings turn over at different rates: the composition of dwellings sold would therefore differ from that of dwellings standing. Countries with large public-housing sectors could have overstated dwelling wealth unless these dwellings are excluded.[1] Similarly, if privately rented dwellings are owned by corporations, not other households, their exclusion could reduce measures of dwelling wealth, especially in countries with low owner-occupation rates.[2]

With these data caveats in mind, Table 1 shows the shares of non-financial assets in total household wealth for countries for which we have sufficient data. Although these data include consumer durables for all countries except New Zealand, non-financial assets are dominated by the dwelling stock. A decade and a half ago, Australia's household balance sheet contained a non-financial asset share around the international average. Since then, the share in most other countries has fallen or stayed fairly constant, while in Australia the share has risen by almost 10 percentage points. This divergence is not due to differences in the relative importance of financial assets: household holdings of financial assets in Australia are not particularly low relative to those in other developed countries. Rather, housing is expensive relative to income in Australia.

Table 1: Non-financial Assets as a Share of Total Assets
Per cent
  1987 1990 1993 1996 1999
Australia 55 68 63 63 64
Canada 47 46 45 43 42
France 61 57 51 50 na
Germany na 67 65 64 60
Italy 53 51 50 48 na
Japan 65 63 57 51 na
UK 55 55 45 42 42
US 39 38 35 32 28
Sweden(a) 49 53 51 47 45
New Zealand 58 61 57 61 60

Note: (a) 1999 data refer to 1998.

Sources: Mylonas, Schich and Wehinger (2000); RBA; RBNZ

The ratio of aggregate dwelling wealth to disposable income is roughly equivalent to the ratio of average dwelling prices to average disposable income; the ratios will only differ to the extent that there is a difference between the number of private-sector dwellings and the number of households (this difference is marginal in Australia).[3] Table 2 shows this measure of dwelling prices is relatively high in Australia, and grew fairly steadily through the 1990s, reaching 378 per cent by late 2000. While some nations (Japan, UK, Sweden) experienced rapid run-ups in dwelling prices, these booms ultimately led to busts, and price-income ratios in those countries returned to levels closer to those in other countries (Henley 1998). Of this group of countries, only New Zealand has followed Australia in experiencing sustained growth in relative housing prices.

Table 2: Housing Wealth as Per Cent to Household Disposable Income
  1980 1985 1990 1995 1998
Australia 248 239 281 303 355
Canada 123 118 129 129
France(a) 172 218 218 227
Germany(a) 331 302 301
Italy 133 170 172 166
Japan(b) 380 397 641 429 381
UK 343 357 361 252 293
US 169 170 173 155 163
Sweden(b) 208 184 245 182 198
New Zealand 185 237 243 278 283

Notes: (a) 1998 data refer to 1997.
(b) Figures refer to non-financial assets which include consumer durables as well as dwellings.

Sources: Bundesbank; Mylonas et al (2000); OECD; RBA; RBNZ

Some increase in dwelling prices should have been expected through the 1990s in Australia. Following financial deregulation, households now enjoy greater access to loan finance for the purchase of dwellings. Reinforcing this trend, the move to low inflation over that period enabled households to service larger mortgages and therefore purchase more expensive homes (Stevens 1997). These factors would be expected to increase demand for dwellings and put upward pressure on dwelling prices. Household indebtedness also increased substantially during this period, reflecting these changes, bringing Australia to around the average level of indebtedness seen in other comparable countries; we discuss these issues in more detail in Section 4. However, these changes do not explain why the increase in dwelling prices since the 1990s has resulted in Australia having relatively more expensive housing than other low-inflation countries. These changes explain the increase in dwelling prices and indebtedness, but not why the price level has increased from around the average to well above international averages.

This divergence in the dwelling wealth-income ratio, if sustained, implies that different countries have different relative prices of housing in the long run. We consider an explanation for this based on unobservable and unexplained differences in preferences for housing to be unsatisfactory, and inconsistent with the evidence on housing quality presented in Table 4. The ranking of countries by dwelling wealth-income ratio does not obviously follow differences in average income, so these variations in the relative price of housing are also not obviously attributable to housing services being either a superior or inferior good.

2.2 The Relative Attractiveness of Dwelling Wealth

Despite the limitations of the data presented in the previous section, they clearly suggest that Australians have concentrated a larger portion of their wealth in housing than their counterparts in other developed economies, and spend a larger proportion of their incomes to purchase a home. The first step in finding the reasons for this result is to establish whether there are government policies or other factors that could have contributed to it. Tax policies such as exclusion from capital gains tax can make owner-occupied dwellings relatively more attractive than other forms of investment, and thus cause over-investment in dwellings. It is also important to assess whether there is a greater revealed preference for dwellings in the sense of their being larger or higher-quality in Australia than elsewhere. Tables 3 and 4 present indicators for these factors for Australia and the other countries for which we have dwelling wealth data.

Table 3: Policies Affecting the Relative Attractiveness of Housing Wealth
  Mortgage interest deductibility Capital gains exemption on family home Share of public housing Memo item: home ownership rates
  Per cent  
Australia No Yes 5.1 70.1
Canada No Yes 1.7 63.7
France Yes(a) Yes 17.0 56.0
Germany Yes Yes 26.0 43.0(b)
Italy Yes(c) Yes 6.0 68.0
Japan No No(d) 7.0 60.3
UK Yes(e) Yes 24.0 69.0
US Yes Yes(f) 1.2 67.4
Sweden Yes No 22.0(g) 56.0
New Zealand No Yes 6.4 71.2
Notes: Data are latest available. See Appendix B for detailed information on sources and reference periods. There are other more targeted policies that encourage homeownership across countries (Miron 2001). Although these policies can vary across countries, their net effect seems less significant because their coverage is generally limited.
(a) Interest is deductible for the first five years. The deduction is equivalent to 25 per cent of the total interest bill, subject to a ceiling based on the date of the contract and age of the building.
(b) West Germany only.
(c) A tax credit of 27 per cent of interest payments is allowed up to a ceiling.
(d) A special deduction of ¥30,00,0000 can be claimed for the principal residence.
(e) Mortgage interest deductible only on the first £30,000 of a mortgage.
(f) Capital gains is theoretically subject to tax. However, any capital gains from the sale of the family home when another dwelling costing at least as much is purchased within two years of the sale is exempt from taxation. A once-in-a-lifetime exclusion of US $125,000 also exists for people over 55 years.
(g) Excludes co-operative sector.
Table 4: Indicators of Housing Quality
  Persons per room
Average existing dwelling size Average new dwelling size   Houses
Detached houses Dwellings with six or more rooms Dwellings built since 1980
  m2 Per cent of stock
Australia 0.6 131.8(c) 185.5   87.6 78.8 63.5 33.7
Canada 0.5 114.0 na   66.4 55.9 75.0(d) na
France 0.7 88.0 102.5   56.2 na 16.6 32.0(e)
Germany(a) 0.5 86.7 101.9   45.6 31.0 11.5 22.0
Italy 0.8 92.3 88.7   na na na na
Japan 0.7 89.6 94.3   na 59.2 na 51.8
UK 0.6 84.0(f) 76.0   80.7(f) 25.6 36.8 13.3
US 0.5 156.5 199.7   66.7 60.6 45.2 25.4
Sweden 0.5 89.8 86.0   45.7 na na 12.0
New Zealand(b) 0.5 132.0 na   83.0 73.0 56.1 na

Notes: Data are latest available. Proportion of houses in dwellings refers to all single-family dwellings including townhouses and terraces. See Appendix B for detailed information on sources and reference periods.
(a) West Germany only. Dwellings built column refers to dwellings built since 1979, not 1980.
(b) Existing dwelling size and detached house data refer to Auckland.
(c) Excludes public housing.
(d) Refers to five or more rooms. The average number of rooms in Canada is six.
(e) Since 1975.
(f) England only.

Owner-occupied housing is tax-advantaged in Australia, but some developed countries apply an even greater range of tax incentives toward home ownership, including deductibility of mortgage interest payments (Table 3). Past theoretical work suggests that deductibility of mortgage interest represents a greater distortion than capital gains tax exemption (Britten-Jones and McKibbin 1989).[4] On this basis, we would expect that if anything, Australia's housing stock is less affected by over-investment than those of some other developed countries.

The quality of the Australian dwelling stock is comparable with that in some other countries. However, Australia has a greater proportion of detached houses, suggesting somewhat more land-intensive housing patterns, and the share of relatively new homes built in the past 20 years is somewhat higher, due to Australia's relatively high population growth (Table 4). Dwellings in Australia appear to be similar in size to those in other non-European developed countries. Although dwellings are larger on average here than in Europe, this is partly because households are larger; the number of persons per room is around the average for developed countries.

The rate of home ownership in Australia is higher than a number of the countries shown in Table 3, but there are many other countries with similar ownership rates, including New Zealand, Finland, Ireland, Greece and Spain (European Parliament 1996). Home ownership tends to increase average housing prices because owner-occupiers internalise the cost of the wear and tear they create in their home, while renters might not fully bear such costs (Henderson and Ioannides 1983). Owner-occupiers therefore require a lower gross return than landlords, and are thus in theory willing to pay more for a given dwelling in the absence of differences in tax treatment.

On the other hand, ownership of private rental properties also attracts favourable tax treatment in many OECD countries (Miron 2001). Negative gearing tax provisions in many countries allow landlords to deduct interest payments against income from other sources if they exceed rental income net of expenses, while tax credits and loan subsidies apply in France (Cardew, Parnell and Randolph 2000). These tax provisions generally ensure that owners of rented dwellings receive the same tax treatment as owners of other commercial properties (Weicher 2000). Since they are common across developed countries, Australia's negative gearing provisions do not represent a relatively greater incentive to invest in rental properties. In the UK by contrast, mortgage interest cannot be deducted against rental or other income (Miron 2001). This may be discouraging the expansion of the existing small private rental sector there.

Although previous studies have found evidence of over-investment in housing in Australia, the evidence that this over-investment is greater than in other countries is weak (Bourassa and Hendershott 1992). Therefore tax policies do not appear to explain the divergence in the stock of housing wealth between Australia and other developed countries; to do so, the incentives for over-investment would have to be stronger here than elsewhere. The concern about over-investment is probably better directed at countries that allow tax deductibility of interest payments on owner-occupiers' mortgages (Mills 1987).

2.3 Dwelling Prices in Different Cities

Figure 1 indicates that city sizes and city-level house prices are related. Although city-specific factors also matter, larger cities usually have higher average housing prices than smaller cities in the same country, even after allowing for variation in incomes, which are also usually higher in larger cities. Then if large cities make up a relatively large share of the population, the national average dwelling price will be higher than if the same population was spread over a larger number of smaller cities.[5]

Figure 1: US and Australian House Prices by City
Per cent to median gross household income
Figure 1: US and Australian House Prices by City

Sources: CBA/HIA for Australia; National Association of Realtors for US

There are a number of possible reasons for this relationship between city size and dwelling prices. Income differentials are clearly important for explaining the high level of dwelling prices in big cities. Mori and Turrini (2000) argue that transport and communication costs encourage higher-skilled workers to concentrate into large urban centres, while Glaeser and Maré (2001) suggest that the observed wage premium paid in larger cities reflects endogenous improvements to human capital arising from lower search costs and greater specialisation. However, this cannot be the whole story, as dwelling prices are high relative to income in larger cities, as well as in absolute terms (Figure 1; Table 5).[6] This may be because housing demand represents an increasing share of expenditure as income increases: preferences may not be homothetic or wealth may increase faster than income.[7] Other reasons include that larger cities offer more amenities and a greater range of job opportunities. In equilibrium, these benefits will be balanced against greater costs, such as congestion, crime and higher housing costs (Gabaix 1999b).

Table 5: Dwelling Prices by City Relative to Disposable Income, 1998/99
City Population Average income Dwelling price-income ratio
  '000 Per cent of national average Disposable income Gross income
Sydney 4,041.4 113.1 8.06 5.64
Melbourne 3,417.2 113.2 4.69 3.51
Brisbane 1,601.4 97.1 5.16 3.92
Perth 1,364.2 100.4 4.87 3.76
Adelaide 1,092.9 91.4 4.21 3.47
Canberra 348.6 124.7 3.83 2.94
Hobart 194.2 93.3 3.38 2.58

Notes: These price-income ratios are not strictly comparable with the national data in Table 2. Survey data understate national accounts disposable income, and number of households does not equal the number of dwellings.

Sources: See Appendix B


The measure of dwelling wealth in household wealth calculations is based upon private dwellings. Since this measure is expressed as a percentage of household disposable income, the inclusion of public housing would inflate this estimate because households do not own them. [1]

Whilst this effect is likely to be small in most countries, it will be more important in Continental Europe (especially France, Germany and Italy) where some rental housing is financed by large businesses. [2]

Holiday homes and vacant rental properties can result in the number of private-sector dwellings exceeding the number of households. [3]

One policy encouraging home ownership in Australia that is not seen elsewhere is the exclusion of owner-occupied dwellings from means and assets tests that can restrict access to government pensions. This encourages pensioners to hold onto larger homes rather than trade down to something smaller, thus restricting the supply of family-sized homes available to larger households. This tax advantage to owner-occupied housing is not applicable in other countries because their welfare systems are not means tested in the same way. [4]

This assumes that there is no systematic negative relationship between house prices in large cities and the share of large cities in the national population. The trend line for US data was estimated by OLS. The coefficient on population is clearly significant using White's correction for heteroskedasticity. If we regress (log) house prices on log income and population, instead of the price-income ratio on population, the coefficient on population remains significant, while the coefficient on income is not significantly different from 1. [5]

As shown in Figure 1, the divergence increased between 1998 and 2000. Canadian housing prices also show a roughly rising relationship, but this is dominated by unusually high housing prices in British Columbia and low prices in Quebec. [6]

It might also be because large cities are space-constrained, limiting supply and putting upward pressure on dwelling prices. [7]