, , ,

Each month, real estate boards release new data.  The media report the press releases because, in part, the headlines are dramatic and readers are curious.  This post goes beyond the headline to show what is easy to find and some of what it means.

Consider this week’s headline on sales through the Toronto Real Estate Board (TREB) during August 2013: “Toronto real estate prices jump with condos helping rise in sales” followed by about 600 words and a couple of bar graphs (Toronto Star) .  The original source of information had 27 pages of mostly numbers.  (TREB is relatively large and has the resources to provide such detailed information.  Other boards provide much less information publicly (see last page), perhaps for a reason noted below.  What extra information is provided and what kinds of questions should be asked?

I choose to focus on two pages (p. 2 and 26 copied into an xlsx file).

Page 2 contains rarely discussed information on the distribution of selling prices.  Even if a headline focuses on the average price, the figure below shows the dramatic difference in the distribution of prices for apartments owned under a condominium contract (red) and detached houses (blue): it is not just the difference in the average but also the spread (i.e. the peak is lower and the right tails of the distribution is much longer).

Selling Prices TO Aug 2013

Page 26 offers a more challenging insight and a little-known innovation in the Toronto area.  A house price index (HPI http://housepriceindex.ca/) controls for observable differences between houses to show how the city’s overall price level rises or falls.  Some very serious intellectual fire power is used to create these indices based on “repeat sales” data .  The method is even more complicated than that used for tax assessment purposes.

The estimated change of 3.7 percent (after adjusting for quality) is much smaller than the Aug. 2013 vs. Aug 2012 change in the unadjusted average price (which is 5.4 percent).  Thus, differences in the types of houses sold this year (i.e. bigger, more valuable locations, better access to transportation, …) explain about 1.7 percentage points of the increase.

The table on the index values shows how the estimated price change becomes more variable at a more disaggregation level.  Even if the change is 3.7 percent for the city as a whole, the estimated change in many zones is less than 0.0 or more than 7.0 percent.  Whether this range is because of local differences or because the estimate becomes less precise is not clear from this table.  Since the sample size is smaller, extra variability should be anticipated.

Random variation makes a journalist’s job easier; there is always some place going up.  A lot.  Even so, don’t expect to see a headline as boring or as honest as “This month’s random variation is 4.76”.