Will AI Replace the Commercial Real Estate Appraiser?

August 6, 2019

 

Will AI Replace the Commercial Real Estate Appraiser?

By Jonathan Rivera, Director, Capright

Stephen Hawking feared that the development of full Artificial Intelligence (AI) would spell the end of the human race. So, with recent exponential advancements in AI, it is natural to think that AI will soon replace the need for a human commercial real estate appraiser. The ability to efficiently collect and aggregate vast amounts of data has greatly improved the quality of Automated Valuation Models (AVMs) – sophisticated computer algorithms that estimate the value of a given property. AVMs have long been in the mainstream for the single-family housing market and now there are a number of groups trying to expand the use of AVMs to commercial properties. A growing percentage of our industry believe that AVM’s data capabilities, efficiency, and non-bias judgment will make them equal or even superior to human appraisers. However, when you consider the limited, non-standardized data sets, the subjective nature of much of the appraisal process, the numerous judgement calls that an appraiser must make, and the regulatory environment that requires personal accountability, it is unlikely AVMs will fully supplant the human appraiser in the near term.

Appraiser Judgement

From market rents, to expense estimates, to investment parameters, appraisers make dozens of opinions throughout the valuation process. While many of these decisions are based directly on hard data, sometimes they can be more nuanced. For example, a recent sale of a newer Class A apartment property with a price that is notably below the asking price may appear to be an indication of a declining market. However, after interviewing brokers active within the immediate market, the appraiser learns that the property had several design flaws that required the buyer to invest significant capital to remedy. Or as another example, when reviewing the characteristics of a vintage apartment tower, on paper the property appears to be a typical older Class B urban high-rise building. However, after further research, the appraiser learns that the building is recognized in the market as a “trophy” asset and would likely sell with a slightly below-average cap rate due to its architectural uniqueness. In these instances, human judgment and observation would take into account important factors that an AVM has no ability to detect.

The AVM is a Black Box

For some, the allure of the AVM is that it would produce a fairly predictable result based solely on aggregated data. This may work fine during a normal upwardly trending market, but as we discovered during the 2008 financial crisis, complex algorithms that few understand can cause significant harm during a sustained downturn. During the early part of the last decade, automated systems were used to build mortgage backed securities with high credit ratings. However, when borrowers began to default, it was difficult for banks to understand which parts of the mortgage backed securities were worthless and which were creditworthy. It would take years to untangle them. Due to the complexity of the underlying computing models, and the limited number of people who understand their mechanics, AVMs for commercial assets have similar risks. In contrast, traditional appraisals are straightforward and all factors relevant to the valuation are explained in the narrative. Furthermore, the client has the ability to talk with the appraiser who is personally accountable for the opinion.

Computers Do Not Buy or Sell Commercial Real Estate

An appraisal is an opinion of value that assumes an arm’s-length sale between a willing buyer and seller. The likely buyers for most of the large commercial real estate assets are financial institutions, pension funds, and investment trusts. As such, the judgments and decisions the appraiser makes about buyer and seller motivations inform the selection of methodology and assumptions employed in the valuation process. As of yet, an AVM is not capable of this kind of human judgment. The computer algorithms used in an AVM are great at accurately and efficiently pulling in data from various sources, but they are unable to grasp the nuanced decision-making process of a likely buyer or seller. At the end of the day, commercial real estate is a relationship business, and it is not possible for an AVM to capture buyer and seller motivations.

AVMs and the Single-Family Housing Market

For several years, AVMs have been in use in the single-family residential (SFR) market. In many cases, these AVMs work well and are more efficient than traditional human appraisers.  This is especially the case when the type of housing type is commoditized (e.g. a large suburban residential subdivision). However, even within the SFR market AVMs have their limits. For a substantial portion of the market, current regulations prohibit AVMs for federally insured loans, or for mortgages with sale prices in excess of $400,000. Furthermore, due to various subjective factors that go into SFR appraisals, lenders recognize that AVMs are more helpful as a starting point rather than the final estimate of value.

The Stakes are Higher for Commercial Real Estate

Commercial real estate assets are often significantly larger, highly unique assets with appraised values that depend on a myriad of complex estimates and assumptions. This level of complexity heightens the likelihood of an error during the valuation process. Just one incorrect assumption can lead to a massive swing in the value estimate. For the time being, humans are still much better than computers at recognizing when “something is off” with the value of a unique asset and are better able to identify errors in assumptions.

While an appraisal does involve a great deal of data analysis and calculation, it still remains more art than a science. Appraisers are good at observing and integrating subjective factors that may not appear in the data. Furthermore, the report and the appraiser are invaluable to client during the appraisal review process when determining how and why certain decisions were made. Finally, human judgment is the most important factor when determining the actions of a likely buyer and seller. This does not mean that AVMs are worthless. Just like other technological advancements, they will likely be an increasingly useful tool as part of the appraisal process. But only humans can produce appraisals. For now, the best an AVM can do is assist in the data collection process and help set general guidelines for appraisers who must assimilate subjective data and make judgement calls for which they are accountable. While AI may spell the eventual end of the human race, it will not extinguish the near term need for human commercial real estate appraisers.