Automated valuation models (AVMs) rely on the accuracy, coverage, and timeliness of the data they use. But in geographic areas with less consistent data available, the accuracy of the estimated property value may suffer.
For example, valuations may be less accurate in more rural areas because of fewer comparable sales and greater diversity in home characteristics, according to a study titled Appraisal Accuracy and Automated Valuation Models in Rural Areas by The Journal of Real Estate Finance and Economics.
High accuracy usually depends on detailed, localized data, which may limit a model’s coverage in less-represented areas. This is the challenge between accuracy and coverage in AVMs, both of which are required for generating property valuations.
Higher accuracy comes at the cost of coverage and vice versa. Finding a balance between the two is the key to building a high-performing AVM that fits its unique requirements and provides real value for its users.
What influences accuracy in automated valuation models?
What influences an AVM the most is the quality, quantity, and diversity of the underlying data. High-quality, consistent, and up-to-date data improves accuracy by ensuring the model captures relevant property features and market trends. The most significant data points are the number of recent home sales and how similar the property is to other properties in the surrounding area.
When excessive variability is present in a dataset, such as diverse home sizes, and economic factors, this lessens accuracy by hindering the model’s ability to find consistent patterns and generalize well.
The more generalized a model is, the less regressivity, or likelihood of performing poorly, the model has. But this does not mean that datasets that are too narrow are the optimal path. They may yield high accuracy for specific cases but also lack the ability to perform well across broader applications.
Other factors to keep in mind that may influence accuracy include the choice of algorithms, features, and validation processes.
The role of coverage in automated valuation models
Coverage in AVMs refers to how many types of homes and areas a model includes. Broader coverage improves model functionality across more locations but can compromise precision for unique properties. Narrower coverage offers better accuracy but is limited to specific areas or home types.
When an AVM lacks coverage, this may be because of several factors, including:
- Limited data availability in certain geographic areas
- Reliance on outdated data
- Insufficient representation of diverse property types
- Lack of local market insights
Strong coverage helps to create applicability across a broad spectrum of properties. However, achieving extensive coverage often requires compromises on the granularity of data, which can affect the accuracy of valuations in specific areas.
Increasing coverage with the use of more generalized data can lead to less accurate valuations, creating a trade-off that developers and engineers must navigate in their AVM designs.
Balancing accuracy and coverage in automated valuation models
Choosing how to balance accuracy and coverage depends on the specific needs of the task. For example, a lender might use a highly accurate AVM for final loan approval but a broader coverage AVM for initial screening of loan applications.
For tasks requiring broad property data representation, such as broad market analysis, coverage is preferable. AVMs with extensive coverage can provide insights into market conditions across different regions.
Likewise, investors rely on accurate property valuations when looking to buy or sell any real estate assets but require broader coverage for investment portfolio valuations. They may use an AVM with broader coverage to include all properties required for a portfolio valuation, but a more accurate AVM for individual valuations.
Another case for accuracy would be when compliance with financial regulations is necessary. Regulatory compliance requires accurate valuations to ensure adherence to legal standards and reduce the risk of penalties.
How Xome approaches the future of its automated valuation models
It is a challenge for AVMs to achieve accurate valuations while being applied broadly across geographic areas and property types. As the use of automated valuation models for property valuations continues to rise, striking a balance between accuracy and coverage will become even more crucial. Finding a unique approach with different models and rules for different property types may help developers gain higher accuracy and coverage in their AVM.
Our engineers at Xome® are constantly working to improve our valuation model to make sure we provide the most accurate results to our customers. With regular model maintenance, updates, monitoring, and human oversight, we can stay on track with our AVM development goals and provide balanced property valuations to our users.
Want to try out the Xome Value®? Search our properties for auction and for sale, or check your home’s value estimate on our new Home Value page.
