Everyone is eager to spot the top of a housing market, but it is difficult to do because the sales mix clouds what is really happening as the market transitions. The sales mix is not represented in the average and median and even the HPI the Van RE folks are so happy to trumpet.
As the market peaks, the speculative buyers gradually fall back and the real buyers (those shopping for shelter, especially former bears capitulating) buy whatever they can afford. The shelter buyers end up in housing of a much lower grade than they would prefer, and they overpay for it. This last part is critical. You can see this happening in Toronto, the <500,000$ sales are the main thing moving. This pulls the average down, as well as the median once the total sales become significant, but it forms a false peak in the average and median because these buyers are still badly overpaying for what they are getting, just the total $ changing hands is lower, pulling the stats down. Still firmly a bubble, but one running out of participants.
Zip ahead a bit. The market fully turns and those with a speculative component to their ownership, and those who have overextended and really want out, rush into the market to sell. Now we get the opposite effect. Those on the sidelines suddenly see bargains (that won't look like it later, but that's another post) and jump on them. The "bargains" they are getting are things like formerly 900,000$ houses that are being fire sold for 700,000$. These bargain hunters are happy to overextend to get into the market. Happy happy! And this repeats itself across the price spectrum, and the average and median are pulled upward by it, when in fact, the market for any given house type is firmly in plummet mode.
HPI, which measures the price of a single house type, will eventually show this peak, but not until stage 2 above. In stage 1 above, the price won't change (or it may inch upward), but sales will decline because buyers are participating in a different tier than the one being tracked. (HPI doesn't report the number of houses that qualified for the calculation in any given month.)
The teranet sales pair measure WILL pick this up, but you won't see the report until 3 months after the fact.
You could probably spot the peak early if you had access to sales data with square footage, and neighborhood quality to run a regression. I don't have access to that data for Canada or Australia.