Blogs & Comment

Measuring house prices

The new repeat sales price index ( RSPI)is major development for Canada, I believe. We will have a less biased idea of how house prices are changing. Yet, there doesnt seem to be much coverage in themedia so far.There has been some in the blogosphere. One in particular stands out, Larry MacDonald – UnderappreciatedGuru(I swear, I do not knowthis person, nor were anyinducements offered!)
Anyway, the RSPI likelyjust needs time to catch on– the index was only announced earlier this month. One thing that could help get the ball rolling is when a financial group announces they have developed futures contracts or other derivatives to trade off the index as was done for the Case/Shiller house price index in the U.S. market. Such derivatives open the door to hedginghouse prices, which was a good ideain the U.S. where house prices are now down nearly 20% over the year — see Trading housing futurespost of Aug. 30, 2006.
While it appears the RSPI is a better alternative than the average-price measure presently used in Canada, some readers did not seem convinced it had solved all the problems. Paul wrote:
if youre looking at repeat sales, whats the chance that these houses were bought, renovated and then flipped for a higher price? This would skew prices higher than normal.
I put this question to the creators of the index and got this response back from Simon Ct, Managing Director of Property Derivatives at National Bank of Canada:
I will spare you all the detailed methodology (you can download the methodology from our website). Basically, the contribution of each transaction pair to the index value is not equal. The weighting of a given pair depends on how close its annual growth rate is to the average annual growth rate. The price of a house that has been subject to significant renovations will display an annual growth rate that is significantly different from the average, and its contribution to the index value will be significantly down weighted. Moreover, if an asset property type changes between the first and the second transaction of a pair, the pair is rejected.
Blogger Michael James on Moneywrote about the RSPI that:
it does shrink the sample space considerably. Also, any method that looks only at sales can give skewed results. For example, if a period of time has mostly low-end houses sold, then looking at sales gives little information about the change in value of high-end houses. The best metric will be one that attempts to value all houses.
The reply from National Bank was:
Sample size: we screen ALL transactions registered. The actual pair count is published on the website (bottom part of the charts). There is more than enough data for the estimation to be statistically robust. A sample too small would result in high short term volatility in the index value. Sales vs. appraisals: we still think actual sales data is more representative than any appraising method (especially in periods of low trading activity). Sometime in the future, we may be able to publish index values for different tiers, high, medium and low price houses, but our first tests show very little difference between the three in terms of annual growth rate.