The Teranet – National Bank House Price Index is estimated by tracking the observed or registered home prices over time. Properties with at least two sales are required in the calculations. Such a "sales pair" measures the increase or decrease of the property value in the period between the sales in a linear fashion. The fundamental assumption of the constant level quality of each property makes possible the index calculation but imposes difficulties in selecting (or filtering) those properties that satisfy it. This difficulty arises from the lack of information about the property, and only the amount of price fluctuations versus time may provide an indication on possible changes in the physical characteristics of the property or non-arms-length transaction. Such properties may not be included in the estimation process.

For the Teranet – National Bank House Price Index, all properties that have been sold at least twice are considered in the calculation of the index; this is known as the repeat sales methodology. Properties that are affected by endogenous factors are not considered in the estimation. These factors may include: a) non-arms-length sale, b) change of type of property, for example after renovations, c) data error, and d) high turnover frequency (biannual or higher).

In the repeat sales methodology, the averaging of price appreciation from different pairs of sales is done using a complex estimation process in which each pair is a separate observation. Click here for a detailed description of the methodology used to calculate the individual city indices as well as the national composite index.

A simplified example

Assuming the example of three houses in the same region that traded during a three year period:

• Sales pair on house A: Sold in 2005 for \$300,000, and in 2006 for \$330,000
• Sales pair on house B: sold in 2005 for \$250,000, and in 2007 for \$297,000
• Sales pair on house C: sold in 2006 for \$280,000, and in 2007 for \$308,000
• With the first pair, we learn that house prices increased by 10% between 2005 and 2006. { \$330,000 / \$300,000 – 1 }

• Assuming that the sales information from house A is correct, we can imply that the value of house B in 2006 was \$275,000. { \$250,000 x ( 1 + 10% ) }

• Between 2006 and 2007, we can do the following inferences:
• From house B, we can infer that the change was 8%    { \$297,000 / \$275,000 - 1 }
• From house C, we can infer that the change was 10%  { \$308,000 / \$280,000 - 1 }

• Thus, from 2006 to 2007, the house price index change is the average of the increase in house B and house C which in this example is 9% { ( 10% + 8% ) / 2 }