RDP 2005-10: Housing and the Household Wealth Portfolio: The Role of Location 2. Literature Review

There is little doubt cities can provide benefits, such as employment and business opportunities, that result in higher income and asset holdings of their residents. The literature on urban economics, which studies the economics of agglomeration, goes back to von Thünen in the 19^th century.^[3] While this strand of literature initially focused on industrial agglomeration and the benefits derived from it, another focus has been on population density and the human capital benefits resulting from frequent contact between many people. This aspect has been especially prominent in literature analysing the earnings gap between urban and rural areas. Benefits such as the development of skills and the imitation and transfer of knowledge that arise from geographical concentration have been well documented in the recent urban economics literature, particularly Glaeser (1999) and Glaeser and Maré (2001). While these agglomeration effects can boost incomes, they also imply a higher cost of housing due to the relative scarcity of land in more densely populated areas (see, for example, Ellis and Andrews 2001, Gramlich 2002 and Voith 1999). A natural result of higher dwelling prices is that it can affect households' choice of asset allocation for those households that decide to own their home rather than rent. Two effects can be at work here. First, to secure access to the higher income (and the associated accumulation of wealth) that comes from city living – the urban wage premium – households should be willing to allocate at least some portion of their extra income/wealth to housing. Exactly how much depends on a number of factors, including the nature of household preferences and the extent to which housing services are divisible.^[4] Ultimately, the way in which the share of assets held in housing responds to variation in income and wealth is an empirical question.

The second effect is due to the possibility of additional (net) benefits that make households willing to allocate a greater share of their assets to housing in cities. These non-pecuniary benefits include greater amenities and lifestyle opportunities typically associated with cities. By itself, this second effect implies that households will be willing to hold a more concentrated portfolio of assets in order to access the benefits of city living.

There has been little previous work using household-level data linking the share of housing in households' assets with urbanisation. Some empirical work has shown that location matters for the housing tenure choice (Curcuru 2003 and Rapaport 1997), but the relevant factors in the tenure decision can have quite different effects from those in the asset allocation decision. Studies that look more generally at what determines household portfolio composition, such as health (Rosen and Wu 2004) or age (Ameriks and Zeldes 2001), typically focus on financial asset portfolios rather than property portfolios. Like other studies on household portfolio composition, such as King and Leape (1998), we focus on those households that own an asset, rather than modelling the participation decision, and we focus on housing assets, rather than net housing wealth, since we want to abstract from the financing decision.^[5]

Micro-data studies that link urbanisation and wealth generally find a significant effect between the two. For example, a study by Fisher and Weber (2004), using data for the United States, finds that people living in non-metropolitan areas are up to 2 per cent more likely to be net worth poor than those living in large metropolitan areas. In a different context, using Swedish data, Goetzmann, Massa and Simonov (2004) find that under-diversification of equity portfolios in urban areas is strongly linked to the professional and geographical proximity of investors to specific stocks. That is, rural investors, who have less intimate knowledge of specific stocks, tend to hold more diversified equity portfolios.

Such micro-data studies on household asset allocation decisions have not been attempted in Australia before, with household-level wealth data only becoming available quite recently with the 2002 HILDA Survey. At a macroeconomic level, Ellis and Andrews (2001) observe that Australians tend to hold more of their assets in housing than households in other countries. They suggest that this is due to the high average level of dwelling prices resulting from the unusual concentration of Australia's population in two large cities.

With disaggregated household asset data now available for Australia, we examine the share of total assets that owner-occupiers are willing to devote to their own home, across different urbanisation levels, controlling for wage-premium effects by means of income and wealth.

Footnotes

For a detailed discussion refer to Fujita and Thisse (2002) or Krugman (1996). [3]

Variation in income/wealth can occur along two dimensions. First, income/wealth can vary within a given location across households (reflecting variation in their earning capacities). If preferences are homothetic and housing services are perfectly divisible, variation in income/wealth should have no effect on the share of income/wealth devoted to housing. However, a positive or negative effect could occur under other types of preferences, or if housing services are sufficiently lumpy. The second dimension to consider is changes in income/wealth across locations for a given household (that is, with unchanged earning capacity and other characteristics). Again, it is easy to show that the effect on the share of income/wealth devoted to housing could be negative or positive depending on preferences and/or the lumpiness of housing services. The key here is that house prices in both locations will be determined endogenously so that in equilibrium the average household is indifferent between locations, assuming constant non-pecuniary benefits. [4]

Portfolio studies, especially of equity holdings, typically also include the rate of return of the asset as an explanatory variable. This is more difficult in the case of housing, since this asset combines an investment and a consumption decision. In the case of the location decision, it may also involve an employment decision, especially if comparable rental accommodation between locations is not readily available (as could be expected in countries with high home ownership rates, such as Australia). [5]