RDP 2018-03: The Effect of Zoning on Housing Prices 3. Decomposing Property Values into Dwelling Structure and Land

To estimate the effect of zoning, we consider the cost of a marginal increase in the number and density of dwellings for a given area and given population (hence a reduction in average household size). This means that we do not include the costs involved in increasing the footprint or size of a city, such as provision of extra roads and utilities, as a cost of supplying housing.

We first decompose property values into contributions from land and structures. We have three main sources of information on this, each of which gives broadly similar results (Table 1). Readers who are not interested in the strengths and limitations or calculation of the various measures can take our baseline estimates as given and skip ahead to the next section. Further details are in the Online Appendix (Section 1), available in the supplementary information provided with this paper on the RBA's website.

Perhaps the simplest and most direct approach to splitting land and structure values is to use estimates of land values from state valuers general, who collect estimates for tax purposes. In Australia, land taxes are levied by state and (especially) local governments. These estimates are constructed using sales of vacant lots and improved properties, with the latter adjusted to remove the value of the improvements.

The NSW Valuer General provides estimates of land values for individual properties in Sydney, which we match to our sample of sales in 2016. For the other three cities, we use valuer general estimates of land values for detached houses by local government area, from which we construct a city-wide average. Estimates are shown in row 1 of Table 1. For example, the estimated average land value for Sydney for detached houses sold in 2016 is $663,000. The $497,000 difference between this and the average Sydney sale price of $1.16 million represents our first estimate of the average value of structures, shown in row 2 of Table 1.

Valuers general dedicate considerable resources to accurately producing these estimates, and conceptually they are closely aligned with our purposes.^[3] However, there is the possibility of downward bias in land value estimates if landowners challenge high estimates of their land value, but not low estimates.^[4] Anecdotal reports suggest this bias may be significant. In addition, there are likely to be some small differences in how the estimates are constructed in different states. One advantage of the valuer general estimates is that they are disaggregated by local government area, which we examine in Section 7.

The ABS's Building Activity Survey (BAS) provides estimates of the value of new detached dwelling structures by state. For example, as shown in row 3 of Table 1, our estimate from the BAS of the average cost of building a new house in NSW in 2016 was $346,000.^[5] However, we need estimates of the value of existing structures, which will tend to be smaller than the value of new structures, given that house size has increased over time. We account for this by adjusting construction costs in proportion. The average house built in NSW in 2016 was 227 square metres, implying an average construction cost of $1,520 per square metre (row 4). We multiply this construction cost by the average floor space of existing houses (194 square metres) and add a 7 per cent upward adjustment for period houses (discussed in the Online Appendix) to estimate the value of existing house structures of about $314,000 (row 5).

A third source of information on construction costs is provided by Rawlinsons Construction Cost Guide (Rawlinsons Group 2017) and Rider Levett Bucknall's (RLB) ‘Riders Digest 2017’ (Rider Levett Bucknall 2017). These are widely used references in the construction industry that provide estimates for different types of construction projects in different cities. Rawlinsons and RLB report estimates on a cost per square metre basis. As shown in row 6, our estimates from these sources suggest that in Sydney a new house would cost $1,810 per square metre to build. As shown in row 7, this would imply that a new Sydney house would cost $412,000 to build, while replacing the average established Sydney house with one the same size would cost $374,000 (row 8). This method is similar to the approach used by Glaeser et al (2005), who use industry-sourced estimates of construction costs to derive estimates of structure value.

Two difficulties in inferring the value of existing dwellings from new construction costs are allowing for depreciation and renovations and dealing with ‘period houses’. We discuss these in the Online Appendix, Section 1.

The three estimates of construction costs discussed above each have their strengths and weaknesses. However, for our purposes the similarity of the estimates is more important. They range from 27 per cent to 43 per cent of the average Sydney sale price, from 31 to 39 per cent for Melbourne, from 39 to 64 per cent for Brisbane, and from 32 to 51 per cent for Perth. None of the three approaches consistently generates the highest or the lowest estimate. Our estimates contain some outliers. For example, the structure value in Sydney based on valuer general estimates is high relative to other cities and other approaches. The industry-based structure estimate in Brisbane is high relative to the other estimates. We expect that closer examination of the underlying data would narrow these ranges, but this would have little effect on our conclusions and is left for further work.

For purposes of a baseline, we take a simple average of the structure value estimates from each method, shown in Table 1. For Sydney, this means an average structure cost of $395,000 and an average land value of $765,000 given the average sale price of $1.16 million.

The estimates in Table 1 can be compared with other estimates. The annual national accounts estimate that the value of residential land in Australia is approximately twice the value of residential dwelling structures. This is similar to the results in Table 1. Unfortunately, however, the national accounts estimates do not distinguish between detached and high-density dwellings. Moreover, estimates are available at the state but not city level. That said, these data can be useful for evaluating national trends (e.g. Lowe 2015).

The UDIA (2017) report that vacant lots of land sold for an average of $465,000 in Sydney and $237,000 in Melbourne in 2016. These estimates are far below the prices of land in Table 1. This is because most undeveloped land is located on the outskirts of cities and these vacant lots tend to be relatively small.

Urbis (2011) estimates development costs for detached housing on greenfield sites and for apartment blocks on infill sites. These estimates have been emphasised by Hsieh, Norman and Orsmond (2012) and RBA (2014). After allowing for inflation and differences in house size, the Urbis construction cost estimates for detached housing are similar to those in Table 1, which is unsurprising given that they are sourced from Rider Levett Bucknall. Urbis estimates of land costs for detached housing are much lower, reflecting their focus on greenfield developments in outlying suburbs (and the rapid growth in land prices subsequent to their study).

Urbis make explicit allowance for some additional costs that we exclude, most importantly subdivision and infrastructure costs. This reflects their different focus. Whereas they are interested in the cost of converting more land to residential use, we are interested in more intensive use of residential land that is already serviced.^[6] A liberalisation of zoning would not directly increase the population or area of a city, so the costs of urban growth are appropriately excluded from our estimates of marginal cost.^[7] In any case, these costs are not large, relative to our estimates, and would not change our conclusions. An upper bound on these costs is provided by Urbis's estimates of average costs per dwelling for Sydney greenfield sites: subdivision construction costs were estimated at $47,000 and infrastructure costs (including Section 94 contributions) were $44,000 in 2011. Together these amounted to about 13 per cent of the average Sydney sale price. The Housing Industry Association (2011) and Centre for International Economics (CIE 2011) argue that a large part of the ‘infrastructure charge’ does not reflect genuine costs. These costs are much lower on infill sites and in other cities. The Online Appendix (Section 1) discusses these costs and charges in more detail.

Footnotes

The NSW Valuer General (Valuation Services 2017) provides a 225 page manual describing procedures for land valuation. [3]

A 2005 report (NSW Ombudsman 2005) documented cases of conservative valuations being applied in some areas, but these issues may have since been addressed. [4]

Industry sources suggest that construction costs are typically higher outside of capital cities. Building approvals data are available at a more disaggregated regional level, but are likely to be more volatile (Hodges 2015). [5]

In technical terms, they hold density or factor proportions constant and vary the scale of production. For an examination of the effect of zoning, it is appropriate to hold the scale constant and vary density. [6]

Of course, increased housing supply in one locality would encourage migration to that area. But changes in zoning policy would not directly affect national population so there would be offsetting infrastructure savings elsewhere. For a given population, average infrastructure costs often decrease as density rises, in which case these ‘costs’ would be negative. [7]