RDP 2018-03: The Effect of Zoning on Housing Prices 6. Decomposing Average House Prices over Time

Figure 2 shows a decomposition of property prices back to 1999. To extend our estimates of structure value back in time, we adjust the estimates from Table 1 by movements in the producer price index (PPI) – output of house construction series for each city's corresponding state, and by movements in the average floor area of houses sold in each city.[17] To estimate the value of physical land over time, we estimate separate regressions for each year back to 1999.

Given that housing supply takes time to respond to developments in demand (even in the absence of zoning restrictions), looking at the zoning effect over a period of a few years may reduce concerns that a gap between prices and costs reflects temporary supply shortages or delays.[18]

In all four cities, our estimate of the zoning effect increased rapidly as a share of average house value over the early 2000s, alongside rapidly rising house prices. Subsequent to this initial rise, the patterns across cities diverged somewhat. The increase in our zoning effect estimates over this time period are likely due to existing zoning restrictions becoming more binding as demand has risen, rather than a meaningful change in the restrictions themselves.

Our estimates suggest that while structure values have increased in all four cities since the early 2000s, they have declined substantially as a share of average house prices. Estimated physical land values have increased more rapidly over time in Melbourne and Sydney than in Brisbane and Perth, but have accounted for a relatively small share of the divergence in average house prices between the four cities over the period.

Figure 2: House Price Decomposition
By city, mean
Figure 2: House Price Decomposition

Sources: ABS; Authors' calculations; CoreLogic

Our data extend earlier in time than shown in Figure 2. However, the contribution of land becomes so small that property values fall within the range of our estimates of construction costs in Brisbane and Melbourne. Estimates for early years would be decomposing measurement error. Indeed, our earliest estimates of the zoning effect in Brisbane and Melbourne are slightly negative. We suspect this arises because the industry-based approach to estimating structure value is too high for these cities. Notwithstanding this, our decomposition suggests that zoning restrictions were not a material constraint on housing development in these cities at this time – unlike in Sydney, where the zoning effect is estimated to have been around 20 per cent of average sale prices, even in 2000. This also shows that our methodology does not result in substantial estimates of the zoning effect in every place and at every point in time.

Footnotes

Changes in these state PPI series are almost identical to changes in the corresponding capital city CPI new dwelling series, but for technical and conceptual reasons are slightly more appropriate for our use here. We also adjust for the introduction of GST. [17]

Although if the housing market (as an asset market) is forward looking, then expectations of a supply response should offset this to a large degree. [18]