NSW Using Regression to Estimate Property Prices

I am not versed in statistics or the 'regression model.. but as a valuer it would be negligible to apportion a value over certain sections of a dwelling/unit to determine the overall valuer of the subject. Yes, there is the hypo, cost to build and summation methods, but these are almost usually a secondary 'check' method and should never be solely relied upon in isolation.

The only apportionment in value I would do for properties is for separate car spaces in the CBD vicinity, which attracts a definite premium. To apportion values in a micro sense, within bedroom, bathroom, aspect and the like is difficult.. what assumption do you place on condition and material and most importantly and near impossible to measure, buyer preference?

 

How does this take into location, permitting rules, ease of consolidation of block, views etc?

 

It doesn't in precise terms. The coefficient values are essentially an average of 100 sales in Surry Hills/Redfern which have occured over the past year. If you consider that a property has something going for it/against it beyond the area/number of beds/baths/car spaces, you can adjust your value, but it probably goes without saying that these characteristics will be the main driver of prices in the area. If I wanted to, I could incorporate additional has a good view (yes/no), close to train station (yes/no), new build/old build etc. and rerun the model to determine if these statistically change the value of a property.

 

Of course, a model is only as good as its inputs and its limitations, they should never be wholly relied upon, but they can be used as a good benchmark. The model is limited by its inputs, but you have to be wary of doing what's known as 'overfitting' a model. However, picking up on your point about accounting for the condition of the property, I had a look at a particular housing market and subjectively made a determination as to whether a property was unrenovated (renovator's delight), had a 'basic', 'average' or 'great' fitout. If your benchmark was unrenovated, then a basic fitout add (on average) $69,145 to the value of a property, an 'average' fitout add (on average) $90,460 to the value of a property and a 'great' fitout add (on average) $129,820 to the value of a property.
Statistically, the value of the properties in this particular market are:
$66/sqm land+$33548*(#bedrooms)+$32261*(#bathrooms)+$10547*(#car spaces)+ [$0 for unrenovated OR $69145 for a basic fitout OR $90460 for an average fitout OR $129820 for a great fitout]

Here is a sample of the input and output... The full database incorporates 102 sales records.