Photo by Markus Spiske temporausch.com on Pexels.com

This article was posted in Realtor Magazine and can be found here. It was written by John S. Brenan who has been the Director of Appraisal Issues for The Appraisal Foundation since 2003.

Because you may be reading this on a laptop, tablet, or smartphone, you already know that today we use technology in ways we never imagined even a few years
ago. Who could have dreamed of ordering something online and having it delivered within hours? Now we’re anticipating deliveries via driverless cars and flying
drones.

With these advances, will computers inevitably replace appraisers when it comes to valuing homes? That question is the subject of much debate. In some limited
transactions, an automated valuation model may be used appropriately today instead of an appraisal. Based on the specifics of the property and the transaction details,
an appraisal may be unnecessary. For example, I’d be irate if I owned a $2 million home free and clear but had to pay a large fee for an appraisal in order to take out
a $50,000 line of credit. However, if I’m looking to buy a $500,000 home with 10 percent down, is it reasonable for a lender to rely on artificial intelligence to
determine whether the collateral is adequate? Not likely.

I couldn’t agree more with the sentiments of Karen Belita, a data scientist with the National Association of REALTORS®, who wrote in a blog post, “When it comes
to online home value estimates, the number one caveat for consumers is that these estimates are not a substitute for formal appraisals, comparative market analyses,
and the in-depth expertise of real estate professionals.” Bravo. Indeed, AVMs are not appraisals. It’s possible that as technology evolves, AVMs may be used to a
greater degree. But today, in many cases, an automated valuation is suspect if there is a lack of available data or the property isn’t a “cookie cutter.” Many of us have
checked our own properties against the finding of an AVM and thought, “Yeah, right.” When it comes to AVMs, your mileage may vary.

So why aren’t automated models more reliable in more transactions? Because computers don’t buy houses; people do. An AVM does a great job of analyzing tangible
features such as a property’s age, number of bedrooms and baths, square footage, and lot size. However, a property’s overall appeal is something that has been, at
least to date, extremely difficult to quantify. It’s a uniquely human phenomenon; a property’s overall appeal reflects a combination of characteristics. While not
everyone has the same preferences, some unusual features will likely face significant market reluctance.
But wait, you say, aren’t appraisers required by the Uniform Standards of Professional Appraisal Practice to be “independent, impartial, and objective”? Absolutely.
Still, appraisers are not machines. They must have relevant data and logic to support their analyses, opinions, and conclusions, but they also incorporate the concept
of market value reflecting the interests of consumers who are “typically motivated” and “well-informed.”

Recognizing that AVMs play a role in developing an appraisal, the authors of USPAP acknowledge their relevance with respect to their use of regression, adaptive
estimation, neural network, expert reasoning, and artificial intelligence. But appraisers remain better than AVMs at recognizing motivations and knowledge levels
of market participants.

The output of an AVM is not, by itself, an appraisal. It may become a basis for one if the appraiser believes the output to be credible for use in a specific assignment. If
the appraiser believes it to be credible. Today, that’s a very big “if.” So unless and until AVMs can better emulate the human factor, an ethical and competent
appraiser remains indispensable.

The Appraiser’s Coalition of Washington along with the Network of State Appraiser Organizations (NSAO) aka “The Network”, have asked that readers of this article share it on their social media accounts, sent it to agents, lenders and AMC’s as well as to State and Federal Legislators.