RDP 2001-08: City Sizes, Housing Costs, and Wealth 1. Introduction

Australia's household sector appears to hold a greater proportion of its wealth in dwellings than do households in other countries. This does not appear to be due to greater quality of the housing stock in Australia than in other comparable countries. Since the measure of dwelling wealth used in household wealth calculations includes all private dwellings, this difference similarly cannot be due to differences in home ownership rates. Alternatively, this pattern in the composition of wealth could be a consequence of public-policy decisions that make the purchase of dwellings relatively more attractive in Australia than in comparable countries.

In this paper, we present evidence against this explanation. We focus instead on an alternative explanation which relies on the observation that housing costs are high in Australia's two largest cities, Sydney and Melbourne. Housing is usually more expensive in large cities than in smaller cities, particularly those cities that dominate other urban regions. The two largest cities in Australia account for a much larger proportion of the total urban population in Australia than is the case in most other developed countries. Therefore, the expensive cities in Australia raise the average level of dwelling prices more than in other countries, resulting in a higher share of wealth concentrated in housing.

This concentration of population in the two largest cities is a result of the unusual structure of Australia's urban population. To a first approximation, the populations of cities in many countries are in inverse proportion to their ranking by population size; that is, the second-largest city is roughly half the size of the largest, the third-largest one-third the size and so on. This empirical regularity is an example of a power law, and is known as the rank-size rule or Zipf's Law. Australia's large towns and cities follow an approximate rank-size power law, but with a flatter distribution than other countries. Sydney and Melbourne together therefore account for ‘too much’ of the urban population of Australia, raising the national average price of dwellings and the share of dwellings in household wealth.

Although this geographic explanation for the structure of Australian households' balance sheets is not the only plausible one, it is consistent with the data and does not rely on presumed differences in preferences. The available data do not suggest that Australians have a greater preference for housing than residents of other developed economies, or that government policy encourages dwelling investment to a greater extent than does policy elsewhere.

The characteristics of housing also support a geographical explanation. Housing differs from the standard neoclassical good; it is heterogenous and its spatial fixity means that the location of the housing stock matters to households (Smith, Rosen and Fallis 1988). Housing is thus imperfectly substitutable across locations (Maclean 1994). These factors, combined with construction lags make housing supply inelastic in the short run, so the housing market is prone to rapid increases in dwelling prices (housing price booms). The relative dominance of the larger cities may therefore also help explain Australia's susceptibility to housing-price booms. In countries with less concentrated urban populations, price booms in one city have less effect on national average prices.

To the extent that average dwelling prices are higher in Australia, some households might respond by reducing their demand for housing services. At the margin, renters would shift their consumption away from housing and downgrade to lower-quality dwellings. However, owner-occupiers' housing demand is both a consumption and investment decision, so their response is less clear (Henderson and Ioannides 1983). It seems likely that any demand substitution would only partially offset the initial increase in dwelling prices.

The question of why Australia's population structure is different remains. The broad similarity with Canada suggests that our federal structure has resulted in state capital cities acting as primate cities dominating the surrounding regions. This will tend to flatten out the rank-size relationship. It is also possible that large countries with small populations – like Canada but unlike the United Kingdom – have flatter rank-size relationships because the large distances between population centres increase transport costs. If so, the structure of the household balance sheet in Australia would not require a policy response, but rather would be partly a necessary implication of our geography and political history.

The paper proceeds as follows. In the next section, we document the importance of dwelling wealth in the household sector's balance sheet in Australia and other developed countries, and critically examine some possible explanations of the high level in Australia. Section 3 provides a brief overview of the literature and empirical evidence about the distribution of city sizes. The data confirm that Zipf's Law is a reasonable first approximation to the distribution of city sizes. We also show that Australia's urban structure accounts for around one-third of the difference between the wealth-income ratios in Australia and the United States.

Section 4 shows that the effects of urban structure on national average housing prices might only occur if households' financial behaviour is not constrained by either financial regulation or the interaction between capital market imperfections and inflation. Section 5 develops a simple model of city sizes with housing costs consistent with the observation that larger cities have more expensive housing. Therefore, the more important are the large cities in the total population, the higher will be the national average level of dwelling costs. The conclusion in Section 6 draws out some of the macroeconomic implications of Australia's relatively large share of dwellings in household wealth, and in particular, argues that a dramatic fall in dwelling prices is unlikely.