RDP 2006-03: Australian House Prices: A Comparison of Hedonic and Repeat-sales Measures 2. House Price Concepts and Definitions

When discussing house prices it is important to clarify whether we are interested in a price measure relating only to those houses sold in a period or a measure of prices relevant to all houses in the economy. If it is the former which is of interest, then concerns regarding compositional and quality change are not relevant. However, in the latter case, these issues need to be given careful consideration.

Specifically, a researcher is often concerned with how the price of a representative house (or the stock of housing) in the economy changes over time, assuming constant quality. This amounts to measuring the pure price change in housing, and is central to understanding the dynamics of housing markets for several reasons: it is the concept that is relevant for household consumption, borrowing and investment decisions, and for analysis of changes in household welfare; it provides information on past returns from a constant-quality investment in housing, which matters for the portfolio decisions of households and firms; and it can be useful in assessing whether house prices reflect fundamentals or are over-or under-valued as an asset (and, more generally, the efficiency of housing markets).

When measuring pure price changes, there are various weighting approaches. If the researcher is interested in the change in the value of the housing stock (or a representative portfolio), then a higher weight should be given to price changes in higher-value houses because of their greater share in the total value of the housing stock (Shiller 1991). On the other hand, if the researcher wishes to measure price changes in the representative house, then an equal weighting of observed house price inflation rates would be appropriate.^[3] For the remainder of this paper, the discussion centres on the pure price changes for a representative house, assuming an equal weighting.

When measuring pure price changes, this should ideally be imputed from the prices of all houses in the economy. However, since houses are heterogenous and traded infrequently,^[4] changes in the composition of properties sold can influence a simple transaction-based measure, such as a median. For instance, if the share of higher quality (and hence more expensive) houses sold rises, a median measure will show an increase in price even if the prices of all houses in the economy did not change. More generally, a transaction-based measure of prices, which does not adjust for compositional change, may not be representative of the price movements relevant to all houses in the economy.

Adjusting for changes in the quality of the housing stock is also required for measuring pure price changes. New house construction and renovations to existing houses tend to improve the quality of the housing stock over time, contributing to higher prices, while depreciation of existing non-renovated houses tends to reduce quality. If the former effect dominates, then a simple transaction-based measure will over-estimate prices over time, reflecting quality improvements as well as pure price changes.

It is difficult to disentangle observed price movements into their respective price-and quality-driven components. Because location is unique, even with identical housing structures, an inherent difference exists in the quality of properties. This can be problematic when measuring price changes over time if, for example, the relative value of any particular location changes. From a practical perspective, it is difficult to obtain comprehensive data on all aspects of quality that influence the price of a property, and to measure changes in quality precisely over time. However, to the extent that data on quality are available, there is scope to come up with better estimates of pure price changes using this information.

Footnotes

Alternatively, when combining observed prices levels of housing sub-markets, for example in the construction of a mix-adjusted measure, turnover weights or (ideally) housing stock weights can be used across market segments. [3]

On average, approximately 6 per cent of the dwelling stock is turned over in any given year. [4]