RDP 2014-06: Is Housing Overvalued? 4. Data Summary

The credibility and usefulness of our estimates hinge on the quality of our data. However, a detailed discussion of measurement issues takes considerable space, which we leave to Appendix A. That discussion includes our treatment of historical estimates. In this section, we provide a brief overview of current estimates.

An important feature of our dataset is that it contains matched prices and rents for the same properties. Rents are obtained by RP Data-Rismark from listings on major real estate websites, which cover almost all properties advertised for rent in major Australian cities. RP Data-Rismark also estimate prices for these same properties, as explained in Appendix A. As a simplification, if the property has been sold within the previous ten years, the sale price is brought up-to-date by multiplying it by the change in a hedonic price index since the time of the sale. In the absence of a recent sale, the price is imputed by detailed hedonic regressions.

For interest rates, we use the average fixed 10-year mortgage rate, less a term premium of 1.3 percentage points. (Like other measurement details, this estimate is discussed in Appendix A.) To express this in real terms, we subtract financial market inflation expectations over the next ten years, derived from swaps markets. We assume that the opportunity cost of owner's equity is close to the real mortgage rate, so leverage can be ignored.

Our estimates of running costs are based on expenses that landlords claim against rental income on their tax returns. The largest of these are council rates, repairs, and expenditure on furnishings and equipment. We exclude expenses that would be paid by investors but not owner-occupiers, such as property agent fees and land tax. We assume the remaining expenses are representative of what owner-occupiers would also expect to pay, on average.

The most important transaction cost is stamp duty, rates for which we obtain from state revenue offices. We base our estimates of other transaction costs (selling commissions, conveyancing, etc) on discussions with industry contacts. We amortise transaction costs over ten years, the median tenure of home ownership.

Our estimates of depreciation come from Stapledon (2007). We measure capital appreciation on a constant-quality net-of-depreciation basis (to be comparable with rental data, in principle), which means the estimates of depreciation have no net effect on the user cost.

In contrast to other studies, we do not explicitly include a ‘risk premium’. It is not clear that the financial risks of home ownership should outweigh renters' insecurity of tenure and uncertainty regarding future rents, or how this might be quantified. Moreover, aversion to risk is just one of many unobserved subjective factors that may influence the decision to buy a house. We prefer to calculate expected returns and allow households to compare these with their own weighting of subjective factors. We do, however, calculate a rental premium in Section 5.3 which encompasses attitudes to risk and other non-financial considerations, and find this to be small, on average.

Whereas the above elements of the user cost can be observed or estimated with some confidence, the expected rate of capital appreciation cannot. A variety of assumptions can be made for this variable, as we discuss in more detail below.