Boosting the Accuracy of Commercial Real Estate Appraisals: An Interpretable Machine Learning Approach

Area of Science:

• Real Estate Economics
• Applied Machine Learning
• Financial Valuation

Background:

• Market valuations in the U.S. commercial real estate (CRE) sector often exhibit inaccuracies and biases.
• Existing valuation practices may not fully capture the complexities influencing property prices.
• The NCREIF Property Index (NPI) provides a dataset for analyzing CRE market valuations from 1997-2021.

Purpose of the Study:

• To assess the accuracy and bias in U.S. commercial real estate market valuations.
• To evaluate the potential of machine learning algorithms, particularly boosting trees, to improve valuation accuracy.
• To extend machine learning applications in property valuation beyond residential and multifamily to office, retail, and industrial CRE assets.

Main Methods:

• Utilized a dataset of U.S. commercial properties from the NCREIF Property Index (NPI) spanning 1997-2021.
• Employed machine learning algorithms, specifically boosting trees, to analyze valuation deviations.
• Incorporated 50 covariates to identify structured variations in the differences between market values and transaction prices.

Main Results:

• Boosting trees successfully captured and explained structured variations in deviations between market values and transaction prices.
• The application of boosting trees led to increased appraisal accuracy and the elimination of structural bias in valuations.
• Model understanding was highest for apartments and industrial properties, followed by office and retail buildings.

Conclusions:

• Supervised machine learning methods, like boosting trees, offer significant potential to enhance state-of-the-art valuation practices in the U.S. CRE sector.
• This study is the first to apply machine learning to office, retail, and industrial CRE valuation, expanding upon prior residential applications.
• Findings are relevant for authorities, banks, insurers, and investment funds seeking to improve real estate investment and risk management.