First some context, to the uninitiated, CoreLogic’s hedonic regression model basically aims to impute the value of individual properties across the country. In other words, instead of simply using a price-based measure the model is an estimate of housing values.

The revised model takes advantage of new data sets and changes to geographic boundaries. There are no differences to the underlying geography (at the macro level) until you get down to more granular areas.

For example, inner-city Melbourne now has 3 SA2s (similar in size to suburbs).

Overall, introducing a new time weighting to the regression is by far the most significant change to the model.

What this does is place greater importance on recent sales and this means trend lines or inflection points in the market – like moving from a positive to a negative – can be spotted earlier than before.

Index numbers have been rebased

However, given that any additional weighting can increase the volatility, we need to be careful with weighting in the index and I think we’ve found a happy medium here.

A revisionary index, rather than keeping the back series static, marks another major change to our model. Unlike the revised model, monthly numbers (within our former index) wouldn’t change going forward.

For example, a number published in, say, August of this year, would effectively remain the August number.

Moving to a revisional index we get a much better handle on the true history of data trends.

By comparison, given that it allows for a more complete back series, the revised model provides for a full complement of housing sales data. The historical series is now updated monthly and we’ll be revising the numbers on a rolling 12-month window.

Drilling down into the regions, more [granular] data sets mean less volatile results within smaller areas. This should make it easier for valuers to interpret the direction of the trend.

A truer history of data trends

While revisions shouldn’t come as a revelation, it would be more surprising if the revisions were major ones and we’re certainly not expecting that. Drilling down into the regions, more [granular] data sets mean less volatile results within smaller areas.

This should make it easier for valuers to interpret the direction of the trend.

It’s important to note that changes to the model won’t necessarily change the level of house values overnight. Think of it this way; looking through a historical lens, the previous index showed the market moved through a trough in February.

By comparison, the new series reveals a trough a month earlier in January which can most probably be attributed to the impact of our new weighting mechanism.

Monthly differences

Admittedly, there is a different measure of value growth over time, but it won’t differ significantly from month to month.

However, given that the monthly measurements differ slightly, measuring how housing values change since the onset of the pandemic through to the end of September will show a larger difference.

That’s because those month-to-month differences do accumulate. As a case in point, under the new model combined capital cities have changed by 16 basis points.

By comparison, this month saw Sydney and Melbourne revise by 0.28% and 0.16% respectively.