Increasing access, equitability, and rigor in the assessment of Neighborhood Mortgage Discrimination

TL;DR

This paper introduces a new method to assess mortgage discrimination in U.S. urban areas, using spatial modeling to better understand how lending practices contribute to health disparities.

Contribution

The novel contribution is a Bayesian spatial model that enables stable estimation of mortgage denial risks in census tracts with sparse data.

Findings

The Bayesian spatial model effectively estimates mortgage denial risk relative to urban-specific averages.

The method allows for stable estimates even in areas with limited data through spatial and non-spatial smoothing.

The approach is publicly accessible, enabling broader research into mortgage discrimination and neighborhood disinvestment.

Abstract

Mortgage discrimination alters the distribution of investment, opportunity, and economic advantage—key contributors of health disparities. Leveraging Home Mortgage Disclosure Act data, we assessed mortgage denial risk in 380 U.S. urban areas. We estimated the risks by census tract–relative to the urban-specific average—using a Bayesian spatial model with conditionally autoregressive distributions fitted with integrated nested Laplace approximation. This approach borrows information through spatial and non-spatial smoothing, resulting in stable estimates in the presence of sparse data. The method, publicly accessible, allows researchers to apply our approach, fostering deeper insights into mortgage lending discrimination and systematic neighborhood disinvestment.

Figures