Rooned. We’ll all be rooned.

There is something about the way real estate trends are reported that borders on the irresponsible. OK. Real estate is the single biggest exposure that Joe and Jill Average have to the economic cycle and people are naturally interested. I get it. But bearing in mind the level of herd behaviour involved in any market, media exaggeration of any figure that can be converted into a sensational headline can only fuel irrational exuberance and despair.

Are sick people ever just driven to hospital by ambulance? No. they are always rushed to hospital. Do real estate markets ever just fall? No they plunge and then reel and then plunge again. But don’t worry. if you hang in long enough they will eventually surge. All these are verbal commentaries on what looks like a pretty orderly statistical process.

The common thread in all these headlines is they use a measure of middle sale price (usually the median). The headline figures actually differ significantly depending on the data source (REIV or APM) and are possibly dodgy. But regardless of which figure you take, it occurs to me, and I am sure I am not the first to make this observation, that there is a potentially huge transaction bias* in these figures.

We are told that the first home buyer’s market is booming. So there are many more sales in the lower price range. So the median or mean sale price would go down, even in a flat market.

It would be nice if some adjustment were made for this. The obvious way to do it is to benchmark the “fair value” of each property before it goes to market, using its features (such as land size, floor space, age and suburb) and some kind of flexible statistical model – say a neural net or random forest (or just a very flexible regression). You then measure whether each property sold for above or below trend and report the average of this figure (say as a %) each month. You could use the previous 12 months of data with a rolling window.

I know that a team of MBA students at AGSM did something like this for McGrath and Partners in Sydney back in about 2001. But as far as I know their index (I am not clear if it became the McGrath APM Index) has never gained any traction with the commentariat.

According to the Herald Sun, Victorian house prices just suffered the biggest drop in 40 years. And now there is great speculation that the federal government will kill the first home buyers grant in their next budget and fear that this will cause the housing market to plunge and reel. If Kev does kill it, and if the first home buyers market does indeed suffer, then we will expect to see many fewer sales in this market segment. And so we will expect  median house prices…… to surge.


* I am not sure if this is the correct term. I know it has another special meaning.

Author: Chris J. Lloyd

Professor of Business Statistics, Melbourne Business School

5 thoughts on “Rooned. We’ll all be rooned.”

  1. Pingback:
  2. Good points. The standard narratives around immigration, first buyer grants (or should we call them grants for sellers-to-first-homebuyers) and low interest rates driving up prices in the lower ranges over the last year, while stockmarket wealth declines hit the top end, are plausible – and supported by median house prices at the suburb level – but not properly quantified, in part for the composition effects you mention.

    There are several ways into the question of what has happened to prices at given housing quality level(s). Some of them:

    A) Get the raw data on house price distribution for a big area and look at it directly.  For Melbourne, the Age publishes transaction volumes by broad categories like ‘inner east’ every sunday – so the volume data must be retained somewhere, presumably by the REIV at least.

    B) Get median prices for smaller areas.  A web search for _reiv trend price graphs _ will bring up a useful tool to graph median house prices by victorian town/suburb for the last 5 years. the charts show that in the year to Q109 median prices in many expensive suburbs like kew and ashburton and brighton have dropped, while median prices in many less expensive suburbs like caroline springs and altona meadows have risen, in some cases by a lot. 

    The pure composition effect (strong demand from first homebuyers) you highlight could plausibly contribute to the decline in median prices in expensive suburbs, though the logic gets a bit less plausible for increase in the lower-priced suburbs, as the FHBs would have to be preferentially buying top-end properties in these suburbs.  So there has most likely been some price range compression as well.

    C) Get a hedonic  series like . Hedonic series estimate the price of some of the underlying dimensions of housing value like size, #BRs, location etc and are therefore not median indices. Hedonic price indices have not dropped as much as median indices, as would be predicted if the main driver of price falls was composition effects.  However, RPdata only publishes single statewide hedonic series, so you can’t tell if 1BR shacks on skid row are getting more expensive compared to mansions on the golden mile. Rpdata does sell data though; maybe they do or could do a data cut on this.


  3. I was under the impression that ABS 6416.0 measured prices in the way you describe.

    The ABS says: “A price index is concerned with measuring pure price change – that is, it is
    concerned with isolating and measuring that element of price change which is not brought about by any change to either the quantity or the quality of the goods or services for which the index is required.

    A representative sample of project home models is selected
    in each city, prices are obtained each quarter and the price movements for each model
    are weighted together. Constant quality is preserved by calculating price movements on
    a matched sample basis (i.e. the price movements between adjacent quarters are based
    on the same models in each quarter). If the specification of an individual model changes
    substantially or a price is unable to be obtained then that model is excluded from the
    calculation of price movement. Adjustments are made to raw prices to compensate for
    any minor changes in specifications.”


Comments are closed.

%d bloggers like this: