Avm Risk? Avm Accuracy Above That Of Appraisals

[Joe Flacco]

The several different valuation options would need to be completed prior to listing and then accuracy of the different options judged by sale price. That is in my opinion the only way to test "accuracy" of several valuation options.

 

[Carolina1]

I think sale price is the only judge of accuracy or true value. Certainly not a automated model.

Makes sense. So, in state like Massachusetts, if an AVM produced accuracy of ~96% within 10% of the sale price, would you consider that accurate?

 

[Joe Flacco]

Makes sense. So, in state like Massachusetts, if an AVM produced accuracy of ~96% within 10% of the sale price, would you consider that accurate?

I don't know if 10% within sale price or a 20% window is considered "accurate" but maybe accurate enough for whatever the expectation is. I think appraisal developed properly should be 96% within 5% of sale price or 10% window.

 

[ucbruin]

Makes sense. So, in state like Massachusetts, if an AVM produced accuracy of ~96% within 10% of the sale price, would you consider that accurate?

How about the term "credible"?

 

[Joe Flacco]

How about the term "credible"?

Credible only applies to appraisal. Different value options can be judged on accuracy, which is what really matters when it comes to collateral risk.

 

[ucbruin]

Credible only applies to appraisal. Different value options can be judged on accuracy, which is what really matters when it comes to collateral risk.

My response was in regards to "appraisal"....

 

[Joe Flacco]

My response was in regards to "appraisal"....

Credible is what appraisal is judged on because it's not possible to test for accuracy without a sale occurring after appraisal. Trying to use AVM as the benchmark for appraisal accuracy is ridiculous.

 

[Joe Flacco]

I would love to participate in a experiment testing appraisal vs AVM. GSE's should hire some skilled appraisers for a year to test which is more "accurate". It would be like Deep Blue vs Garry Kasparov. Desktop appraisal vs AVM before a listing and judged by sale price. That would be fun.

 

[George Hatch]

A valuation model - including the GSE grids that the appraisers use - can't be any more "accurate" than the buyers in the market who are making the decisions that are going into the database being used. We're supposed to be trying to emulate them, not instruct them. The indulgence in the idea that valuation models (and appraisers) should be expected to be more accurate than the market itself is unrealistic. And arrogant.

You test and calibrate your model by comparing those attempts to emulate and predict the market against the market itself.

 

[Joe Flacco]

A valuation model - including the GSE grids that the appraisers use - can't be any more "accurate" than the buyers in the market who are making the decisions that are going into the database being used. We're supposed to be trying to emulate them, not instruct them. The indulgence in the idea that valuation models (and appraisers) should be expected to be more accurate than the market itself is unrealistic. And arrogant.

You test and calibrate your model by comparing those attempts to emulate and predict the market against the market itself.

Exactly. Sale price is the only benchmark for "accurate".