Learning from Online Appraisal Information and Housing Prices

Abstract

This paper studies how housing prices react to an information shock due to the update in the most popular home valuation algorithm. I find that housing prices react gradually to the information shock: while the reaction in the first month is small, a 1% increase in the online appraisal leads to a 0.65% increase in sale prices 12 months after the shock. The magnitude of the reaction in a ZIP Code increases with its average trading volume over the sample period. In ZIP Codes with top 5% average trading volume, I find that a 1% increase in the online valuation leads to a 1.6% increase in sale prices 12 months after the shock, suggesting overreaction to the information shock. Findings in this paper indicate that machine-learning-based online valuations causally affect housing market transaction prices.