CHAPTER 5 HOUSE PRICE INDEXES
5.1 This chapter compares various approaches used in measuring changes in house prices, including the approach chosen for the HPI. The application of price index theory (Chapter 4) to the calculation of the HPI is detailed in Chapter 11, with supporting detail contained in Chapters 6 to 10.
5.2 House price indexes measure the inflation or deflation of the price of houses over a period of time. The indexes describe the change in prices between a specific period and a base period, allowing comparisons of price movements between time periods and also between geographical regions.
5.3 While the purpose of the HPI is to provide an accurate measure of the contemporary rate of change in the prices of the stock of established houses in the eight capital cities, other price indexes are constructed to meet different user requirements. A number of other measures of housing-related prices compiled by the ABS are published in 6416.0 (Table 5). These indexes are discussed in Chapter 13.
5.4 In Australia, a range of other house price indexes are compiled by industry bodies and economic research firms. The methodologies and conceptual approaches used to construct these indexes differ according to factors such as the intended use of the index, the sources of the house price data used, and the resources available to collect and process the house price data.
5.5 Users should familiarise themselves with the stated purpose of their index of choice to ensure that it is suitable for their particular requirements. Familiarity with the different methodologies used to construct the indexes should also inform the user's choice. An overview of some typical approaches to index construction is provided below.
5.6 The standard procedure for constructing price indexes is to select a sample of representative items and to re-price the identical items through time (a matched sample). This approach is not viable in the case of established houses as the observable prices in each period invariably relate to a different set of houses. In other words, the sales prices observed for any pair of consecutive quarters are completely independent datasets.
5.7 While there is not yet an international standard for constructing housing price indexes, the OECD-IMF Workshop on 'Real Estate Price Indexes', held in Paris in 2006, provided an opportunity for developed nations, particularly from Europe, to document and present their current practices in this realm (Diewert, 2006)(footnote 1) .
5.8 Additionally, a discussion of different methods of constructing real estate price indexes, and an overview of the constraints experienced by national statistical offices in constructing representative measures is contained in the International Monetary Fund’s (IMF’s) Financial Soundness Indicators: Compilation Guide (2006) - Chapter Nine: Real Estate Price Indices. The following extract from that guide (paragraph 9.4) is instructive:
‘Constructing representative real estate price indices is challenging. Difficulties can arise because real estate markets are heterogeneous, both within and across countries, and illiquid. There may be no unambiguous market price. Moreover, such diversity and lack of standardization results in the need to gather a wide range of data to compile indices that are characteristic of the various market segments; this would contribute to high data collection costs and may require greater technical sophistication. Representative real estate prices in residential and commercial markets can be hard to measure accurately given the small samples that are often available, as there may be disparate prices for apparently similar properties and prices may be volatile.’ (IMF, 2006, p. 101)
COMPARISON OF CONCEPTUAL APPROACHES
5.9 The central issue is how to utilise prices for an essentially heterogeneous set of dwellings to construct measures of price change for homogeneous sets of dwellings, and compile and release the results in a timely manner. There are three general approaches that might be used to achieve this: hedonics; repeat sales; and stratification.
Hedonic price indexes
5.10 The hedonic approach views products (such as dwellings) as bundles of characteristics that are not individually priced, as the consumer buys the bundle as a single package. Through the use of regression techniques, the objective is to "unbundle" the characteristics to estimate how much they contribute to the total price.
5.11 There are several ways in which this approach can be employed in practice. A hedonic technique has recently been introduced with respect to pricing computers in the 15th Series CPI. Details of the methods used for computers are set out in Information Paper: The Introduction of Hedonic Price Indexes for Personal Computers, 2005 (cat. no. 6458.0). A similar approach could be adopted for housing although, of course, the type and number of price-determining characteristics would be different. A feature of the hedonics approach is that it generally makes use of more price data than other approaches.
5.12 The effectiveness of hedonics is critically dependent on the availability of data on the price-determining characteristics. Analysis by the ABS has shown that the single most price-determining characteristic is location, followed by an indication of the socio-economic conditions of the area(footnote 2) , and the physical characteristics of the dwelling (such as outer-wall construction, overall size, number of rooms, number of bathrooms). While various characteristics are included in datasets for some cities, sufficient data is not available for timely production of the index using hedonic methods (see Chapter 8 for more information on timeliness of data sources). It would be necessary to collect considerable amounts of supplementary data. Due to the cost of collecting and processing this data, the ABS considers that the hedonic approach is not viable at this time.
5.13 The repeat sales approach controls for compositional change by maintaining a historical record of property sales. When properties are sold repeatedly over time, price changes between successive sales dates are calculated. Regression techniques are used to calculate the overall price index for each quarter.
5.14 To be effective, this approach requires a long time series of price data for individual properties, given their infrequent turnover. As the methodology is premised on the assumption that the ‘quality’ of the individual properties is constant over time, this approach may be more suited to some property types than others (e.g. units), or require supplementary information on property renovations. The nature of the estimation technique also means that at least the tail end of the series is subject to potentially significant revision. At the present time, the ABS does not have ready access to the data required for this method.
5.15 The stratification approach involves grouping the observations for the ‘most similar’ dwellings into clusters, to enable the derivation of a representative sale price for the cluster (usually the median price). The objective is to optimise the physical homogeneity of dwellings within each cluster, while ensuring a sufficient number of observations to produce a reliable median price.
5.16 The effectiveness of the stratification approach is determined by the degree of stratification possible and the availability of stratification variables. It may not be feasible to employ fine level stratification if there are insufficient observations to produce reliable movements for each cluster.
5.17 Given the absence of a comprehensive national dataset to enable the use of either the hedonics or repeat sales approaches, the only option currently available in practice to construct an HPI which controls for compositional effects is the stratification approach. While stratification can deal with compositional effects, it will not adjust for quality changes such as the size of the dwellings increasing over time.
5.18 Chapter 11 provides more detail on the practical application of the stratification method.
5.19 Further insight into the merits and limitations of various methodologies for calculating house price movements in the Australian context can be obtained from Hansen (2006).
1 The Paris workshop indentified the need for an international manual or handbook of methods for real estate price indexes. A manual is under development and scheduled for release by the OECD in mid 2011. <back
2 The analysis used the ABS' Socio-Economic Index for Areas (SEIFA) which ranks geographic areas according to their social and economic conditions. For further information, refer to
Information Paper: An Introduction to Socio-Economic Indexes for Areas (SEIFA), 2006 (cat. no. 2039.0). <back
This page last updated 11 December 2009