How does algorithmic trust affect pricing decisions? Evidence from the housing market

Abstract

In November 2021, Zillow announced the closure of its iBuyer business. Popular media largely attributed it to a failure of its proprietary sales price forecasting algorithm, marking one of the most prominent instances of failure in artificial intelligence (AI) based algorithms in the field. We study the effect of this event on consumers’ trust in another algorithm developed by Zillow, the Zestimate, which estimates the current sale price of properties. Our findings show that following the iBuyer closure, home sellers’ list pricing decisions deviated more from the Zestimate values, suggesting a loss of trust in the Zestimate algorithm post-iBuyer closure. Notably, home sellers deviated more by increasing rather than decreasing their list prices. Surprisingly, this translates into properties sold for more and in less time in our empirical context. We conclude the paper by discussing broader implications for other consumer-facing AI algorithms.

Keywords: Algorithmic Pricing, AI Trust, Economics of AI, Housing Market, Zillow