RDP 2014-09: Predicting Dwelling Prices with Consideration of the Sales Mechanism 2. Data and Measurement
September 2014 – ISSN 1320-7229 (Print), ISSN 1448-5109 (Online)
- Download the Paper 822KB
Our primary data source is a near-census of all dwelling sales in Sydney and Melbourne between March 1993 and December 2012, which make up about 40 per cent of all sales in the Australian housing market over that period. These data are provided by Australian Property Monitors (APM),[3] and are an update of data previously used by Prasad and Richards (2008) and Hansen (2009).
Private-treaty is the most common mechanism used for selling dwellings in these two cities. Sales where an auction mechanism was used (or planned to be used) as part of a successful sale make up around 12 per cent of the Sydney sample and 17 per cent for Melbourne (Table 1, columns one and two).
| Transaction type | Percentage of total observations(a) | Percentage of observations filtered for analysis(b) | |||
|---|---|---|---|---|---|
| Sydney | Melbourne | Sydney | Melbourne | ||
| Pre- or post-auction | 2.73 | 3.72 | na | na | |
| Sold at auction | 8.83 | 13.01 | 9.30 | 13.90 | |
| Private treaty | 88.46 | 83.26 | 90.70 | 86.10 | |
| Auction frequency | 11.56 | 16.73 | 9.30 | 13.90 | |
| Total observations | 1,763,032 | 1,677,925 | 1,652,585 | 1,498,549 | |
| Notes: (a) Percentage of total observations where an auction was used (or
planned to be used) as part of a successful sale (b) Percentage of observations after removing identified pre- and post-auction sales, private-treaty sales where an auction was used in the 90 days prior to the exchange of contracts, and observations where prices are not disclosed or there are address inconsistencies |
|||||
In the analysis that follows (Table 1, columns three and four), we restrict our attention to properties sold successfully at auction when measuring auction prices. When measuring private-treaty prices, only those properties sold directly via a bilateral negotiation, with no involvement of an auction in the selling process, are used.[4] Using hedonic price regressions similar to those discussed below, the average conditional price difference between a property sold through an auction and through a private-treaty is 4.2 per cent for Sydney and 5.1 per cent for Melbourne.[5]
To measure average prices we use hedonic price regressions. At the city-wide level, Hansen (2009) has shown that hedonic regressions can provide an accurate estimate of the composition-adjusted price change in dwellings – that is, average price growth after adjusting for changes in the mix of dwellings sold. The specification we use has the general form:
The variable lnPijt is the logarithm of the sale price for dwelling i, in postcode j and at time t; Dit is a time dummy equal to 1 if sold in quarter t and zero otherwise; PCij is a postcode dummy equal to 1 if dwelling i is located in postcode j and zero otherwise; and Cikt is the measure of the kth characteristic (or hedonic) control relating to the attributes of the dwelling at time t.
For Sydney, the hedonic controls include the number of bedrooms, number of bathrooms and the logarithm of a measure of the size of the dwelling.[6] We also allow for interaction effects between each of these characteristics and the type of the dwelling sold (for example, house, semi-detached, terrace, townhouse, cottage, villa, unit, apartment, duplex, studio).[7] For Melbourne, there are only limited data available on characteristics prior to the December quarter of 1997. To avoid an otherwise substantial reduction in sample size, we omit the bedroom, bathroom and size controls, but include controls for the dwelling type. Similar results are found when including the additional characteristic controls but using a smaller sample that begins in the December quarter of 1997.
Figure 1 reports, for Sydney and Melbourne, two-quarter-ended annualised growth of separate hedonic price indices for auction prices, private-treaty prices and all-sales prices. Figure 1 highlights that there are cycles in price growth over the sample period and that all three measures are highly correlated. However, it also clear that the price cycles are not fully synchronised, with some evidence to suggest that auction price changes lead the dwelling price cycle. This is most noticeable around turning points in price growth in both Sydney and Melbourne.
Footnotes
In providing these data, APM relies on a number of external sources. These include the NSW Department of Finance and Services for property sales data in Sydney and the State of Victoria for property sales data in Melbourne. For more information about these data, see the Copyright and Disclaimer notices at the end of this paper. [3]
See Table 1, Note (b). [4]
This is measured using an additional dummy variable for whether the dwelling is sold via auction or private-treaty. [5]
In the case of a house, the size is the total land area in square metres. In the case of a unit or apartment, it is typically a measure of the building area, but can also be the internal area in square metres depending on the source of the data. [6]
The exception is when comparing forecasts out-of-sample. Given limited attributes data in the early part of the sample, and to avoid an otherwise substantial reduction in sample size, we exclude the bedroom, bathroom and size controls when estimating recursively and comparing out-of-sample forecasts. [7]