Heterogeneous asset pricing: An examination of the Australian residential real estate market

Abstract

This thesis assesses the methods for pricing heterogeneous assets, with specific examination of the Australian residential real estate market. Each of the three primary methods for controlling for heterogeneity in residential real estate pricing research – median-price analysis, hedonic regression and repeat-sales regression – is susceptible to bias. This thesis represents the first broad study of the impact of these alternative methodologies on research outcomes. Methodological comparisons are made in the study of residential real estate market efficiency in Chapter 4. Using an extensive database of property sale and attribute data, indices are independently estimated following the three alternative methods, from which market efficiency, in its weakest form, is tested. Specifically, the autoregressive predictability and seasonality of returns is measured using Box-Jenkins methodology and regression analysis. Access to more comprehensive data than has previously been available to researchers enables the estimation of indices following the three alternative methodologies. This allows a controlled cross-methodology comparison of results and represents a major extension to the previous work into residential real estate market efficiency (Case and Shiller, 1989; Rosenthal, 2006). Furthermore, this study represents the first major study of weak-form efficiency and seasonality in returns to the major Australian property markets of Sydney and Melbourne. The results indicate that a predictable autoregressive lag structure exists in the returns, thus rejecting the hypothesis of market efficiency in the Australian residential real estate market in its weakest form. There is, however, no deterministic monthly seasonality in returns although the results from the alternative indices are not consistent. Median-price index returns, for example, exhibit significant heterogeneity-induced seasonality and negative first-order autocorrelation. This contrasts with the returns to the repeat-sales index, which exhibit significant positive first-order autocorrelation. Chapter 5 examines the effect of sample selectivity in the presence of asset heterogeneity on hedonic pricing models. A substantive literature has demonstrated the existence of a price impact to properties sold by auction relative to those that sell by private treaty (Dotzour et al., 1998; Lusht, 1996; Mayer, 1998). The work presented in this thesis extends the research in several ways. Firstly, sample selectivity between sale methods is tested and controlled for using the Heckman two-stage procedure and a matched sampling technique. Secondly, a larger sample of sales than has previously been considered is fitted to a more completely specified hedonic function. This is made possible by access to a database containing the virtual population of sales from the Sydney house market. Finally, this represents the most thorough study of the impact of auctions on prices in the Australian market. Using selectivity-corrected regression analysis, it is found that the auction sale mechanism has no effect on prices in the Sydney house market. This runs contrary to the results of past research in the Australian and New Zealand residential real estate markets which document a price premium (Dotzour et al., 1998; Lusht, 1996). Unadjusted hedonic-regression analysis is also performed, the results of which support an auction premium. This difference in results demonstrates the selectivity bias which has influenced past research, and a further methodological issue in pricing heterogeneous assets. The pricing and performance of new properties is considered in Chapter 6. This is a further examination of the ability of hedonic pricing models to counter sample selectivity. Sample selectivity-controlled methods are applied to a comprehensive sample of sales from the Perth residential real estate market. Differences in attributes between the sample of new and existing properties are observed. After controlling for this sample selectivity, the results indicate a significant price premium to new houses and units of around 10% and 7%, respectively. An examination of the returns to new properties at their first subsequent sale, however, demonstrates significant underperformance of new residential real estate assets relative to the market. This is the first study to empirically and theoretically assess the pricing of new residential real estate assets. Finally, Chapter 7 applies alternative index estimation methodologies to a study of the value of public and private information in the residential real estate market. Heterogeneity in the market is a major source of information asymmetry. Indices for the Sydney residential property market are estimated from lead sources of information, both public (newspaper reporting and advertisements) and private (agents’ sales records). Correlation and regression analysis demonstrates that private information is a strong predictor of movements in an index estimated from the population of sales, in line with a priori expectations. This is the first study to assess the value of information in this market and represents a new direction for residential real estate research, particularly as housing derivatives markets continue to evolve.