Abstract
Appraisal bias has emerged as one of the most controversial issues in the mortgage industry. In this paper, we examine the effects of historical redlining on automated home valuations (AVMS) and prices paid by homebuyers in predominately Black neighborhoods. Proponents of AVMs argue that reduce biases and errors committed by human appraisers, while critics caution that these algorithms simply reproduce the same biases that are embedded in human-driven decision systems. Based on an analysis of 2018 home sale transactions in Atlanta and Home Owners Loan Corporation (HOLC) ‘redlining’ maps, we find that both historic redlining and neighborhood race have a significant impact on current home values and prices paid by homebuyers. Specifically, holding other factors constant, home prices in historic redlined neighborhoods are associated with a 49.3% undervaluation. Similarly, all else equal, compared with majority-white neighborhoods, home prices in majority-Black neighborhoods are 64.2 % lower. Based on these findings, we propose policy actions to address the disparities between house prices and automated valuation model estimates, such as questioning the appropriateness of the sales comparison approach for determining a home’s value, developing regulations for quality control standards for AVMs, and increasing financing for renovation and small business investment in formerly redlined areas. These recommendations can help account for the effects of past discrimination and ensure that Black communities and their residents accrue similar benefits from homeownership as their White counterparts
Journal of the Center for Policy Analysis and Research, Vol. 2024, 34-47