Thursday, March 15, 2018

Replicating the RBA's housing analysis

Last week the RBA released a research paper which sought to unravel the potential effect of planning controls, like zoning, on home prices in Australia. I think the results of their analysis are wrongly interpreted to be due to zoning, and I quickly made my views known on Twitter.
But, like the good researcher I am, I wanted to check their method. So I put together a sample of land sales from my area and replicated the method, just be 100% certain I understood. Lo and behold, I get the same result. Using twenty-nine neighbouring land sales and estimating ln(p) = A + B ln(area) + e, I get the following result.

The coefficient of 0.54 for ln(area) indicates that a 1% increase in land area only increases the total land price by 0.54%. So if the average price was $100/sqm for a 1,000sqm block (total price of $100,000), an extra 10sqm (1%) would cost just $540, or only $54/sqm.

This indicates that indeed, because of zoning constraints, people are unable to assemble marginal pieces of land at $54/sqm and thus must pay on average $100/sqm instead. The zoning effect clearly accounts for 46% of land value. Quite a stark result.

In my data the actual average land price of a lot was 0.1 pence per square metre while the marginal price was just 0.052 pence.


Yes, my average lot size was 4 acres, 1 rood, and 34 perches and sold at 19 pounds, 6 shillings and 9 pence. My randomly selected sample of sales occurred in December 1851 in South Brisbane. Zoning was still nearly a century away from being invented, and the population of the whole state of Queensland was less than 17,000 people or about half the normal attendance at a Broncos football match.

Let me be clear. Either it is true that the method used by the RBA does identify zoning effects, and therefore also identifies zoning effects of a similar scale 167 years ago in a remote and deserted convict colony that we know did not have planning controls. Or, it is true that the method does not identify zoning effects.

You decide.

*Here is a look at some of the data used for this post.


  1. Better thought of as a subdivision effect and zoning can indirectly create subdivision effects

  2. Nice trolling of a troll-worthy paper.

    Only some land development costs relate to lot area (e.g. lot clearing), others relate to lot frontage (e.g. infrastructure mains) and others are essentially fixed costs per lot (e.g. infrastructure connections and infrastructure charges). It makes no sense to expect per square metre prices to be constant, even for a new subdivision; the developer knows that they cannot reassign land from large to small lots without cost, even when the subdivision is still on the drawing board.

    This Queensland government practice note on ‘Integrated residential Development’, in Appendix B, gives cost breakdowns for lots of various sizes:

    Lots of around 600 sqm cost about 50% more to develop than lots that are 300 sqm, not 100% more. 300 sqm lots cost about 40% more to develop than 150 sqm lots.

    Those arguing that per sqm prices would be equal in the absence of zoning are demonstrating their ignorance about the relevant production process, not their ability to apply sophisticated economic analysis to identify regulatory burdens.

    However, the example of a new subdivision doesn’t fully capture how extreme and detached from reality the RBA paper’s argument is—which in the Brisbane context is essentially that, if not for zoning, developers could buy up houses worth $550,000 and produce new $115,000 lots. Of course, subdivision is allowed in a lot of contexts and does happen—although it’s certainly not a free-for-all. But subdivision produces lots at prices which reflect the cost of a) buying houses at the market rate (not some fantasy notion of what the RBA thinks they are really worth to their owners or what it would cost if you could scrape together dreg bits of land and fashion them into a plot as if residential development were akin to pottery), b) demolishing or moving the existing structure, c) professional fees and construction costs for infrastructure augmentation, d) infrastructure charges, e) time costs, selling costs, and developer profit.

    Where it constitutes market-feasible development (inner suburban areas), it produces lots which are marginally cheaper than existing lots—as they need to be because they are smaller. (It does not produce $115,000 lots because it is very difficult for developers to make a profit when they buy something for $550,000, spend $100-200,000, and then sell for $230-345,000.) It gives people who cannot afford a house on a lot of land in the inner suburbs an additional option. It produces a trickle of supply that reduces the extent to which the city spreads into more remote and less desirable locations, which would increase the locational premium of existing suburbs, but it plays a minor role in this relative to alternative forms of infill development, especially apartments.

    If subdivision were allowed in more contexts it could have a somewhat bigger impact. It is something that I would support, based on real, marginal benefits, not the headline-grabbing but essentially meaningful analysis produced by the RBA.

    If the RBA were really interested is this issue, it could have commissioned an expert study into development costs for low-rise, small-scale infill development in Australian cities, including subdivision, which has received little attention compared with greenfield houses and large-scale apartment buildings. But the starting point of any such analysis would be the reality that, in Sydney for example, the development process would start with acquiring an existing house costing around $1 million, not the RBA fantasy of $275,000—and that’s 25km from the CBD. And any competent analysis would show that it is hard to produce new affordable housing given this reality. And attention would inevitably return to how we got to this point, and the RBA’s role in getting us here, a subject that the RBA has become desperate to avoid.

    1. ...headline-grabbing but essentially *meaningless* analysis produced by the RBA.

    2. Spot on! Thanks for the detailed comment.