Photo:
Getty
Countless hours of work are put into traditional property valuations because they’re essential for more than just valuing a property. Valuations determine taxes, insurance, financing options and more.
Property value typically means the present value of the benefits (in the case of commercial real estate, that is cash flows, or rents) that the property will provide to the owner. These benefits are typically long-term. Valuations, therefore, account for many data points like the quality of local schools and the presence of amenities such as supermarkets, public transit, etc. For commercial property, they might also account for macro-economic data, like current interest rates.
Traditional appraisal methods that account for some (or all) of those data points include:
• Sales comparison: This values a property based on the sales price of similar properties (called “comps”).
• Cost approach: This assesses the cost of each element of a property, and adds the value of the land it’s on. Building cost and land cost are calculated separately and then combined.
• Income approach: For commercial properties, the income approach uses the annual property income to calculate its value by dividing income by the capitalization rate for that market. The cap rate represents the average of local and recent commercial property transactions.
• Discounted cash flow (DCF) approach: Also for commercial properties, this approach models 10 years of income and expenses for a property, “discounting” those future cash flows back to now using a discount rate.
These appraisal methods utilize snippets out of large amounts of data and require many hours, and lots of judgment, from an appraiser.
What Is An Automated Valuation Model (AVM)?
Automated valuation models use mammoth amounts of data, typically a combination of property records, property listings from platforms like Zillow and data indicating the attractiveness of the area. AVMs typically use advanced analytics, such as machine-learning models, to analyze many different data points for a given property to predict a property’s current or future value.
Residential AVMs typically analyze public records to calculate the current value of a residential property. The AVM program runs a regression or machine-learning algorithm that accounts for the home’s size, number of rooms, home quality characteristics (granite countertops, air conditioning, pool, etc.) and location. The result of all that data is typically combined with the property’s price history (for how much did it sell most recently?). The final result is an estimate of the home’s value for a requested date (typically, present-day). Many firms offer this, though Zillow’s Zestimate is perhaps the most well-known example of a residential automated valuation model.
With commercial property, AVMs stand to benefit a wide range of important but labor-intensive processes: preliminary valuations, underwriting, portfolio valuations, assessments of collateral when borrowers become delinquent, risk management and more. Much like with residential real estate, an automated valuation model in commercial real estate represents a set of algorithms that combine inputs (the property’s age, the number of schools nearby or amenities) to calculate property value and account for cash flow. For now, our firm is one of the only providers of AVMs for commercial use; however, two other companies that are trying to transform the way CRE valuations are done are Bowery Valuation and Skyline AI.
How AVMs Address The Challenges Of Traditional Valuation Methods
For decades (maybe even centuries) the basics of calculating the value of property have been the same: Compare the property to other, similar transactions (comps) in the area, where the “art” is in selecting the right comps. Prudent appraisers and underwriters may include a few more metrics or methods, but with traditional valuations, this is really all there is to generating the market value of a building.
Like any other manual process, traditional valuations are subject to human error. Our company’s recent research found that traditional valuations have high error rates — up to 16% for buildings worth $1 million or less. Manual, human processes also come with bias. For example, comp selection is often unconsciously influenced by how familiar a valuation expert is with a property or area. More importantly, comp selection might be influenced by the client of the appraiser (for borrowing purposes, a client wants the valuation to be high; for tax purposes, the client wants the valuation to be low).
Investors and lenders are also conscious of the time lag in seeking a third-party valuation. The delay between ordering a valuation and receiving a report is often three to four weeks.
By employing an AVM, however, the process takes a few seconds and requires no manual effort. With less manual effort, there is tons of potential for time savings for users. Less manual effort means lower risk for human error. We have found that the absolute error of the automated model is below 4% for homes and below 6% for commercial properties, which is much less than the error rates of traditional appraisals.
Some submarkets are too small for a traditional appraisal to properly assess, but with an AVM, there are enough data points to run comparable property reports. But perhaps of the most value to users, an AVM is objective. A valuation based on data increases the valuation’s accuracy and makes it a more reliable choice for investors or lenders.
More Data, More Accurate
Appraisers, investors and lenders can leverage AVMs to get a more accurate value for their property of interest — whether residential or commercial — by compiling much larger amounts of data in much less time than a traditional valuation.