Frank Newman: Valuations and statistics

A few months ago I reviewed two websites that had made finding property data whole lot easier: Trademe Property Insights (trademe.co.nz/property/insights) and Homes.co.nz.

Both apply a secret algorithm, to arrive at a suggested market value for residential properties, and provide a range of information about the property. They take what's called a desk top valuation approach - they look at recent sales in the area and compare the actual selling price of each property against its most recent capital value (as shown in the rates notice). They then apply that percentage change across all properties in the area to arrive at an estimated market value. No property inspection is carried out so the approach is inherently flawed. Both sites are upfront about these limitations and point it out in various disclaimers.

In my review I compared the derived market values from the Trademe and Homes sites for two sample properties to see how close they were to each other. A real estate agent in Tauranga has taken my analysis one step further and compared the prices of 15 properties sold at auction, against the average value obtained from the two websites. That's a valuable insight because it shows how close each is to the actual market value.

Here is a quick summary of the results. 

• In seven of the 15 sales the auction sale price was higher than the average estimated value, and lower in eight cases. In other words, the actual sale value was broadly centred around the values estimated by Trademe and Homes.
• In 10 of the 15 sales the average valuation was within 10% of the actual auction price. That's pretty good as it is not uncommon for registered valuers to differ by at least 10%. The most extreme variance was 33% higher at auction than assessed by the two sites ($1m vs. an auction price of $1.5m).

This comparison shows that when both sites are taken together and averaged, the value is a reasonable approximation of market value. There are situations however when it's not, and those variations can be significant, as was seen in the sample.

That supports the view that desktop valuations are OK for a general assessment, but should be treated with care when dealing with a specific property. Every property is unique and there may be aspects that are not captured in a desk-top approach.

The question that is not answered by the analysis is whether a valuation done by a registered valuer (who does inspect the property) would have been any better at predicting the auction value. That would be interesting, given the huge range of values that supposedly independent valuers acting without bias come up with for the same property.

In my view the notion that a property has an exact value is unrealistic. At best a property would have a value range within 10%, and a willing-buyer willing-seller sale would sit somewhere within it. In this respect buyers and sellers sometimes do themselves no favours if they miss out on a deal for the sake of a few thousand dollars. Sure, even a thousand dollars has worthwhile spending power, but experienced investors appreciate that value is very subjective with there are swings and roundabouts.