RDP 2020-02: The Distributional Effects of Monetary Policy: Evidence from Local Housing Markets Appendix B: Kendall and Tulip
February 2020
- Download the Paper 1,879KB
To determine the proportion of value determined by land (physical land – the marginal value of land – and ‘zoning’ effects) in an area, we estimate a hedonic index for each local area (SA3) and impute the proportion of value that can be attributed to structure and land. This approach is the same used in Kendall and Tulip (2018). It assumes that the value of a property can be broken into a land component and a structure component.
The Land component can be further decomposed into a physical land and zoning effect component.
Using the same unit record data from CoreLogic on housing sales, we take all housing sales that meet the following criteria:
- sold in 2015 or 2016; we do not use sales in the early 1990s due to data limitations in building approvals
- land area less than 8093.7 square metres; as used in Kendall and Tulip (2018)
- for the period and SA3 region, the property is not below the 1st percentile or above the 99th percentile in sales price or land area.
For each SA3 area we run the following regression:
where j represents a dwelling in suburb s within an SA3; ${\tau}_{t}$ represent time fixed effects; and ${\alpha}_{i}$ represent suburb fixed effects.
To determine the value of structure for each region we impute the average real cost per square metre of houses from building approvals data over the ten years to 2016. These data are sourced from the ABS building approvals data. For areas where more than half the houses sold were built on or before 1950, we again follow Kendall and Tulip (2018) and add a 40 per cent premium to the calculated structure value to account for differentials in the value of various period houses relative to modern houses.
The land component of an area's housing value is determined by the mean sales price minus the calculated value of structure for the area
The value of the physical component of land determined by the structure is the mean sales price multiplied by ${\beta}_{1}$ from Equation (B1). The value of land attributed to ‘zoning’ is the residual value of land not attributed to the physical component.
The proportions are obtained by dividing each component by the mean sales price. We include all proportions in the model selection procedures of Section 4.3. And we only use the proportion of the physical land in Section 5.3, as we find that this variable is significant in our regressions in Table 2.