RDP 2014-13: Mortgage-related Financial Difficulties: Evidence from Australian Micro-level Data Appendix B: Hedonic Dwelling Price Adjustment

Matthew Read, Chris Stewart and Gianni La Cava

November 2014 – ISSN 1320-7229 (Print), ISSN 1448-5109 (Online)

The hedonic adjustment method generates estimates of the mean price of dwellings sold for each postcode j and month t conditional on the characteristics of dwellings sold. Following Hansen (2009) and Windsor et al (2014), each capital city is treated as a separate market with hedonic models estimated independently for Sydney, Melbourne and Brisbane. The postcode-time dummy hedonic adjustment model is given by Equation B1:

where Sijt is the sale price of dwelling i in postcode j and month t. The vector of explanatory variables, xit , contains dwelling characteristics, including the (log) area in square metres, and dummy variables for the number of bedrooms and the sales mechanism.[16],[17] All variables are interacted with the property type, allowing their effects to differ between houses and units. The dummy variable Dijt is equal to one if dwelling i is sold in postcode j and month t and is equal to zero otherwise. From the coefficients on these dummy variables (λjt), we calculate the average value of dwellings sold within a postcode in a given month after controlling for observable characteristics. The results of estimating Equation B1 separately for Sydney, Melbourne and Brisbane are shown in Table B1.

Table B1: Hedonic Model Coefficient Estimates for Logged Sale Price
Characteristic City
Sydney Melbourne Brisbane
Unit −0.72*** 0.23** 0.25***
Bedrooms × property type
2 beds × house 0.26*** 0.47*** 0.34***
2 beds × unit 0.38*** 0.44*** 0.37***
3 beds × house 0.46*** 0.69*** 0.50***
3 beds × unit 0.79*** 0.77*** 0.74***
4 beds × house 0.72*** 0.91*** 0.73***
4 beds × unit 1.19*** 1.00*** 1.00***
5 beds × house 0.90*** 1.08*** 0.91***
5 beds × unit 1.25*** 0.99*** 0.97***
6 beds × house 0.99*** 1.16*** 0.98***
6 beds × unit 1.41*** 1.12*** 0.93***
7 beds × house 1.05*** 1.10*** 1.03***
7 beds × unit 1.63*** 0.73*** na
Sales mechanism × property type
Private treaty × house −0.06*** −0.06*** −0.10***
Private treaty × unit 0.00 0.05*** 0.02
Area × property type
Area × house 0.02** 0.07*** 0.12***
Area × unit 0.04*** −0.02 0.01
Constant 12.69*** 11.81 *** 11.65***
Number of observations 601,958 669,398 219,120
R2 0.81 0.82 0.72

Notes: Coefficients of postcode-time dummies omitted; ***, ** and * indicate statistical significance at the 1, 5 and 10 per cent level, respectively; standard errors clustered at the postcode level

Sources: Authors’ calculations; APM

Footnotes

We do not include the number of bathrooms as an explanatory variable, as this information is largely missing before 2005. [16]

Genesove and Hansen (2014) find evidence to suggest that average prices of dwellings sold at auction incorporate information about the underlying trend in prices more quickly than average prices of dwellings sold via private treaty. Inclusion of the sales mechanism as an explanatory variable may also control for other systematic differences in the characteristics of dwellings sold via auction (relative to those sold via private treaty) that are unobserved or otherwise omitted from the model. [17]