From regions to neighborhoods: Algorithmic lending and the rescaling of racial inequality in U.S. mortgage markets

Source abstract

Over the past decade, mortgage lending in the United States has increasingly shifted from locally embedded underwriting practices to algorithmically mediated risk assessment systems deployed most prominently by financial technology (fintech) lenders. While existing research evaluates whether these systems expand credit access or alter average racial disparities, far less attention has been paid to how they reshape the spatial organization of inequality. Drawing on loan-level data from the Home Mortgage Disclosure Act (2018–2020), this study examines whether fintech lending is associated with changes in the geographic distribution of Black-White disparities in mortgage approval and pricing across and within metropolitan areas. Multilevel models reveal substantial compression in cross-metropolitan variation in racial disparities under fintech lending, with reductions of approximately two-thirds in approval disparities and three-quarters in pricing disparities relative to traditional lenders. At the same time, within-metropolitan patterns diverge by outcome: pricing disparities exhibit similar tract-level dispersion across lending regimes, while approval disparities display greater neighborhood-level heterogeneity under fintech lending. These findings are consistent with a rescaling of racial inequality from regionally differentiated patterns toward more localized intra-metropolitan variation. By foregrounding scale as an analytic dimension of financial governance, this study illustrates how algorithmic lending infrastructures may reorganize the spatial levels at which racial inequality is produced and observed in contemporary housing markets.