Algorithmic Outputs as Information Source: The Effects of Zestimates on Home Prices and Racial Bias in the Housing Market

Abstract

This paper investigates market participants' reactions to predictive algorithms and the effects of this public information source on market outcomes. In particular, I study the extent to which buyers and sellers rely on the Zestimate, Zillow's estimate of a home's market value, as well as the interactions between the Zestimate and other information sources. Using detailed property transaction data for 120,482 properties sold between May 2017 and May 2019 in the Greater Philadelphia Area, I show that the sale price of a property does respond to exogenous shocks to its estimated home value. I develop a theoretical framework and provide empirical evidence to show how people use the Zestimate as a source of publicly available information that plays an important role in coordination and helping people reach an agreement. The results suggest that market participants tend to rely more on this public information when it is harder to reach a consensus based on private information. Moreover, I show that people's reliance on the Zestimate might mitigate racial disparities in the housing market by providing less biased information.