Some commercial real estate investors have hesitated to embrace automated valuation models (AVMs), lagging behind their counterparts in the residential market. The reluctance often stems from a lack of trust in the accuracy of these models or a limited understanding of how they function. Despite AVMs not being a new concept in commercial real estate, they continue to ignite debates—especially when weighed against the tried-and-true method of in-person appraisals.
An automated valuation model (AVM) is a software tool that merges sophisticated mathematical or statistical modeling with extensive databases of property values and transaction histories to estimate real estate values. By analyzing comparable properties within the same timeframe, AVMs quickly deliver valuations. Powered by proprietary algorithms, these models generate detailed reports for lenders and real estate agents in seconds, providing a fast and efficient alternative to traditional methods.
These models often incorporate a ‘hedonic’ model—a form of statistical regression analysis—alongside a repeat sales index. By carefully weighing and analyzing these components, AVMs generate property price estimates. In addition to these elements, AVMs typically factor in the tax assessor’s valuation, detailed property information, and a thorough analysis of recent sales of comparable properties.
Automated valuation models are used in various ways in the real estate industry, including to support mortgage underwriting and home equity loans and to aid in loss mitigation and credit risk management activities. Insurance companies use them to determine the value of properties for insurance coverage. Government agencies use them to assess property values for taxation purposes. Leading AVM providers include Zillow, CoreLogic, VeroVALUE, and Equifax.
Despite their widespread use today, automated valuation models in real estate have some drawbacks. For the models to work well, they need high-quality data in sufficient quantity. That’s where the most significant vulnerabilities lie in AVMs. The most cited downside is that they do not and cannot factor in the actual property’s condition when determining value. The models simply assume an average state, which may not be accurate. No matter how precise the model is, it can’t note details or variations in the property’s condition.
AVMs are great at making comparisons, but there can be issues if there’s a lack of comps or transactional data. Due to this, newly built properties can be complicated to value. Automated valuation models tend to be literal and can lack the imagination to create something to serve as comps. AVMs also work based on known factors like historical records, so they often miss intangibles that raise or lower values.
Many in the real estate industry know this, and they’re in search of building a better AVM. Green Street, the independent commercial real estate research and advisory firm, built an AVM that wanted to solve the most significant challenges. The first step was creating a model using three accepted valuation approaches: NOI capitalization, value extrapolation, and sales comps.
Green Street’s AVM is also fed with the firm’s high-quality data sets to improve accuracy. Lastly, the AVM is transparent, and Green Street shows what’s inside the ‘black box.’ Green Street claims their AVM is proving accurate, with single-digit error rates for individual assets and low-digit error rates for portfolios of 25 or more. Most importantly, the company’s clients can see precisely how the AVM arrived at the valuation estimates.
Like many major AVM providers, Green Street is quick to laud its model’s accuracy and comprehensive coverage. While Green Street claims its automated valuation model is one of the most accurate for real estate investors, the industry has to be wary of other AVM challenges, including looming regulatory changes that govern automated valuation models.
On June 24th, six federal agencies finalized a rule establishing quality control standards for automated valuation models. This regulation introduces five key quality control factors to ensure that companies using these models produce accurate and free-from-bias results. Under the new rule, financial institutions involved in transactions with secured real estate mortgages must implement robust policies and procedures to prevent conflicts of interest, safeguard against data manipulation, comply with anti-discrimination laws, and maintain high confidence in their AVM-generated valuations.
The new federal regulations aim to inspire confidence in the credibility of the valuations these models produce. “As with models more generally, there are increasing concerns about the potential for [automated models] to produce property estimates that reflect discriminatory bias, such as by replicating systemic inaccuracies and historical patterns of discrimination,” the regulators’ filing said. Automated models can potentially reduce bias, but it’s also possible biases could be built into the models themselves, thereby amplifying harm because the AVMs process large volumes of valuations. The final rule takes effect at the start of the first quarter of 2025.
The advantages of automated models in estimating commercial real estate values are evident and similar to those of any computerized system. AVMs save time, money, and effort, making countless calculations in seconds. This lowers the cost of valuing properties, and automated valuation models are instrumental in assessing the worth of entire real estate portfolios.
While AVMs offer a helpful way to value properties quickly, they’re not a replacement for a traditional (and human) appraiser. Most in the real estate industry use AVM reports as a starting point to provide a rough value range, supporting the final report from an appraiser.
In some cases, appraisal waivers are applied when lenders feel they have enough information on the property’s current value to waive the need for an assessment from a licensed or certified appraiser. These waivers have existed since the late 1990s, and the FHFA expanded the waiver program in 2017. The housing boom in 2021, when interest rates hit rock bottom, flooded appraisers’ desks with requests. There was already an ongoing shortage of licensed appraisers, so waiver programs expanded again.
Appraisal waivers and heavy reliance on automated valuation models come with risks. Property value inaccuracies could increase as more loans bypass traditional appraisals, leading to higher loan-to-value ratios and increased default risk. Without a strong foundation in property valuations, trust could erode in the mortgage industry, potentially impacting the entire housing market.
AVMs can also make valuation errors with stark consequences. Instances have been reported where automated models inflate values based on wrong assumptions. These errors may even be linked to racial bias by undervaluing neighborhoods with more minorities. The recently proposed federal regulations for AVM quality control address potential biases.
A recent Cityscape: A Journal of Policy Development and Research study examined the potential for AVM racial bias. The researchers didn’t find systematic undervaluation bias, but they did observe that the studied AVM had a more significant percentage magnitude of errors in majority Black neighborhoods than white ones. This could be because the automated model reflected human appraiser bias and historical discrimination patterns.
AVM use continues to grow as the tools become more sophisticated with the help of artificial intelligence, but real estate firms still need to be cautious about how much they rely on them. Automated models have not yet supplanted human valuation estimates on a broad scale, nor should they. Because of accuracy concerns, viewing results from multiple AVMs is the best way to increase confidence in reports and use them to supplement a licensed appraiser’s work. Despite being around for a long time, automated valuation models still carry risks that must be carefully considered.
