Follow Datanami:
October 5, 2017

Machine Learning Used to Value Real Estate


Determining the value of commercial real estate, among the largest of asset classes for investors, remains difficult, likely lending itself to the application of new machine-based valuation models that among other things take advantage of “hyper-local” information about real estate locations.

In a paper published in June, Dutch researchers argued that an “automated valuation model” outperformed traditional real estate appraisals, with an error rate of about 9 percent. The model also brings with it the ability to “produce an instant [property] value at every moment in time, at a very low cost.” the researchers said in a paper title: “Big Data in Real Estate? From Manual Appraisal to Automated Valuation.”

The researchers see an opening for applying big data analytics to divine the worth of one of the largest asset classes. “Determining the value of commercial real estate remains elusively

hard, with a workforce of 74,000 appraisers in the U.S. alone still manually assessing the value of assets sometimes worth billions of dollars,” the researchers noted.

Current appraisal techniques are often “anchored” to earlier valuations or a previous sale price, failing to take into account factors like transportation improvements and other amenities that can boost property values. The result, the researchers argued, is that current appraisal techniques “typically lag the market…with values that are artificially low in bull markets and high in bear markets.”

As with other sectors, the real estate industry has seen a huge influx of data. Hence, industry analysts are just now beginning to merge property data with machine learning techniques. Among the first applications was Zillow’s “Zestimate” valuation model.

Along with a faster, cheaper and more accurate valuation model, the Dutch researchers stressed the growing availability of “hyper-local” information ranging from economic and demographic data gleaned from census roles to social media data, police records and amenities like schools, stores and access to public transportation.

Investors as well as lenders would be among the key beneficiaries of timely valuations that more accurately reflect conditions in a local real estate market. “Precise, timely estimations of property values are critical for real estate investors and lenders to make informed underwriting decisions, where systematic errors or biases in valuations may have adverse effects on the provision of equity or debt,” the researchers noted.

The new valuation model also went beyond traditional property valuation techniques such as capitalization rates currently used to compare real estate investments. Instead, their model incorporated all transactions in a given market, taking into account the “relationship between value and independent variables that is consistent across locations.”

According to earlier research, the difference between traditional real estate appraisals and transaction prices ranges from 10 to 15 percent. Appraisals typically take about three weeks and can cost up to $5,000 for commercial properties.

The Dutch researcher claimed their machine-learning valuation model reduced the error rate of 9.3 percent.

Given increased market volatility, the researchers asserted: “For underwriting and refinancing purposes, automated valuation models can provide an instant indication of property value, which saves both portfolio investors and lenders, as well as those interested in a single property, significant time and resources.”

Recent items:

Inside the Zestimate: Data Science at Zillow

Under Contract: Crowd-Sourced Zillow Algorithm