RDP 2006-03: Australian House Prices: A Comparison of Hedonic and Repeat-sales Measures 5. Conclusion

This paper has summarised some of the key theoretical issues in measuring house prices and how different approaches to price measurement attempt to control for compositional and quality change. In principle, hedonic and repeat-sales approaches can provide the best estimates of pure house price changes as they can adjust for both effects. Mix-adjusted measures also have the potential to be useful in circumstances where changes in the quality of the dwelling stock are relatively minor, which is likely to be the case over short horizons.

Notwithstanding these theoretical advantages, whether the regression-based measures offer a better estimate of pure price changes than a simple median or mix-adjusted measure is largely an empirical question. To examine this I compare several regression-based alternatives using unit record sales for approximately half of the Australian market, between March 1993 and September 2005. The results suggest that hedonic and repeat-sales measures are not overly sensitive to alternative specifications and provide similar estimates of pure price changes. These measures are also comparable to a simple mix-adjusted approach developed by Prasad and Richards (2006). In contrast, the estimates are quite different from the price changes observed in the median, suggesting that compositional change does matter in Australia.

A caveat to these results is that the basic hedonic measure does not appear to capture changes in the quality of houses over time. Using additional quality regressors in the hedonic approach seems to suggest that, if anything, quality may have declined marginally over time (or there has been some compositional shift to smaller properties). This result could reflect the fact that the sub-samples which have most data on housing characteristics are relatively small, and that many aspects of quality are not captured in the available data (for example, quality related to fittings, fixtures, building materials, size of the dwelling structure and so forth). Indeed, using a repeat-sales measure with a constant as an alternative control for quality infers a modest quality-related price increase over the sample period.

In view of the desirable theoretical and practical properties of hedonic and repeat-sales measures, they appear to complement existing house price measures. They provide useful estimates of pure price changes, and information about the magnitude of compositional and quality change in Australia.