Can artificial intelligence finally provide a fair and efficient appraisal process?

Assessors and appraisers from state and local governments often cite data analysis as a double-edged sword when it comes to their day-to-day work. They recognize the value of data analysis but also acknowledge how challenging it can be when the data they need isn’t readily available.

Whether it is assessors focused on conducting property tax assessments across a community or neighborhood, or appraisers focused on single properties, the ability to rapidly draw insights and patterns from massive volumes of data is crucial. And it is why assessors and appraisers are increasingly seeking to leverage artificial intelligence (AI) and machine learning (ML) to deliver a fair and efficient process and valuation to buyers.

Improving property appraisal accuracy with AI/ML
Property appraisal is an optimal use case for AI/ML. There is a vast amount of data available, and ample computing power to process it. Advanced machine learning algorithms can also be used to analyze the data and make decisions. With these conditions, AI and ML can make the property appraisal process easier, faster and more accurate.

One example of how the technology can work in property appraisal is by using a machine learning model that runs thousands of decision trees to compare every property that has sold within a specified window of time. This is known as a comparative market analysis. Rather than having only a few comparisons (or “comps”), every property has as many comps as houses sold in a given period. The model determines how influential every characteristic is, such as location or square footage. It is designed to be completely objective, determining the importance of factors and values of characteristics solely on what it gleans from the data.

In the past, a programmer would instruct a program with rules like, “If a house has one fireplace, increase its value by $1,500.” With AI, the computer processes millions of pieces of property data and, through a series of thousands of decision trees, isolates the value of a fireplace. It learns the value from the data without input from a person, and it determines the specific value of the fireplace for each home, given its geography, size, age or any number of related factors.

Government appraisers can access a website that hosts the output of the property valuation model. In an example model, one report would show the accuracy and precision, which gives appraisers confidence in the model’s algorithms. Another compares the AI value to the working value generated by the district’s traditional method, revealing properties that may have been previously undervalued or overvalued. This allows for a more accurate and fair assessment of property values.

Another advantage of using AI/ML in property valuation is the ability to reassess every residential property in a community every single day. Not only is this a feat that could not be accomplished by humans, but it is done with an exponentially higher level of accuracy and precision. This can be especially beneficial for rapidly changing markets, such as those in urban areas, where property values can fluctuate quickly. Having the day-to-day valuations give assessors insight into market forces at both a high and granular level.

One of the strengths of AI/ML is its ability to identify patterns within data. This can help an assessor identify outliers, such as properties that may have been improved without permits. Perhaps most intriguing is the ability of AI/ML to detect patterns that reveal human bias that is influencing valuations.

Reducing bias with AI/ML
The issue of housing and property market discrimination and bias has thrust itself into the spotlight over the past several months. High-profile reporting of racial discrimination in appraisal valuation was punctuated with the story of a Baltimore minority-owned home being appraised for far less than they anticipated. When re-appraised with a white stand-in as the homeowner and all evidence of a Black family removed, it was valued nearly 60 percent higher.

Bias that favors one demographic group over another is not unique to the real estate industry. The impact of human bias on segments of society—whether deliberate or accidental—has severe economic, social and quality-of-life consequences. Public assessors and private appraisers can use AI/ML to help reduce or eliminate human bias. One of the sources of bias stems from the data that is fed into the valuation process. AI/ML can help ferret out the data sources that are resulting in biased output. Also, because AI/ML uses every other property as a comparable in its assessments, the machine can objectively identify when a property’s valuation is not in line with comparable homes. One report by Brookings Institute found that Black neighborhoods were associated with much lower property values overall, and only some of this was explained by physical characteristics and neighborhood amenities. According to the Brookings Institute findings, median home values in majority Black census tracts were 55 percent lower than median home values in non-Latino or Hispanic white census tracts.

Another type of bias exists when the owners of expensive homes are under-taxed, and the homeowners of modest homes are over-taxed. In other words, a valuation process may overvalue a modest home, resulting in unfairly high taxes. The price-related differential is a statistic used to measure whether high-value properties and low-value properties are assessed at the same ratio to market value. The International Association of Assessing Officers (IAAO) Standard on Ratio Studies suggests that the price-related differential should fall between 0.98 and 1.03. With daily recalculation of this ratio, an assessment office can ensure that it fairly assesses properties of all sizes.

There are many factors leading to appraisal and assessment bias, and it is an obvious problem with a less clear magic bullet solution. One approach seeking to identify and address this bias is using responsible AI to, for example, shed light on potential bias in property appraisals by objectively identifying when a property’s valuation is not in line with comparable homes.

As an example of this AI in action, government tax agencies are using this technology approach in Wake and Mecklenburg counties in North Carolina to track property values for hundreds of thousands of properties on a daily basis. The systems consider dozens of inputs that affect property value: square footage, township, neighborhood, number of bathrooms, exterior finish, or, as the Baltimore family had hoped, renovations.

The system provides tax assessors with an objective cross-check on home valuations, supporting transparent and fair tax assessments. For appraisers, those cross-checks could reveal when bias, either intentional or not, may be creeping in.

 AI is a powerful and valuable tool that state and local government agencies and appraisers across the nation are looking to for improved decisions and efficiencies throughout their operations. Thanks to advancements in AI, instead of being tasked with countless hours of cross-referencing value estimates from a spreadsheet, tools powered by machine learning can immediately and intuitively compare values across an entire state, region, county or city, delivering results faster and more accurately than ever before.

Jennifer Robinson is the leader of global and local government solutions for analytics provider SAS. She has served on the town council of Cary, N.C., since 1999 and is a member of several boards that promote the success of governments, including the N.C. Council of Government and a regional transit authority. She just completed a term as president of North Carolina’s League of Municipalities and is currently serving as the vice president of the National Association of Regional Councils. She has a particular interest in cities using information technology to improve the lives of their citizens and focuses much of her efforts on fostering the use of analytics in creating smart cities.